New Scientist - 08.08.2026.pdf

《新科学家》周刊 2026年8月8日

AI的下一次飞跃

当AI开始体验真实世界时会发生什么？我们即将揭晓

在我们的DNA中发现了幽灵祖先的基因组

数千人接受了具有争议的阿尔茨海默病手术

欧洲如何适应即将到来的山火时代

此外还有：流浪黑洞 / 理解悲伤发作 / 空调的更环保替代方案

No3607 £7.95 CANS11.99

---

麦基伦癌症支持中心

“如果我没有坚持寻求答案，我现在可能已经不在人世了。” —— Denzil

数月来，Denzil一直遭受严重的腹痛，却找不到原因。当检测显示血小板计数较低时，他的全科医生认为这可能是由于他的牙买加血统所致。他通过多次就诊坚持寻求答案，直到最终确诊——结肠癌。从第一天起，麦基伦就提供了建议、指引和支持。但并非每个人都能获得应有的护理。你是否能获得护理，可能取决于你居住的地方、你的收入，或者仅仅取决于你是谁。今天就向麦基伦捐款，您可以帮助确保我们的福利权利顾问、伙伴和支持热线能够为每个需要的人提供服务。您愿意捐赠£1万，以确保没有人会在缺乏所需护理的情况下面对癌症吗？癌症是不公平的，但癌症护理应该是公平的。现在请在 macmillan.org.uk / news 捐款。

扫描此二维码立即捐款

麦基伦癌症支持中心，在英格兰和威尔士 (261017)、苏格兰 (SC039907) 和马恩岛 (604) 注册的慈善机构。也在北爱尔兰开展活动。

MAC20415_NewScientist1 在筹款监管机构注册

---

本期内容

封面故事

30 AI的下一次飞跃

卷 271 第 3607 期 封面图片：Andrea De Santis

7 在我们的DNA中发现了幽灵祖先的基因组 6 数千人接受了具有争议的阿尔茨海默病手术 4 欧洲如何适应即将到来的山火时代 9 流浪黑洞 40 理解悲伤发作 13 空调的更环保替代方案 36 特写：“微生物组分析可能会揭示关于一名未知凶手的线索”

新闻

8 亚马逊的失落文明

一个古老的社会可能曾拥有多达 300 万 人口

1万 量子优势

量子计算机正开始成为有用的科学工具

12 逆转失明

一种可部分恢复视力的眼睛植入物获准使用

观点

18 思考之食

Helen Thomson 谈超加工食品与大脑

19 团块宇宙

Leah Crane 询问：现在是否到了放弃宇宙学中最核心定律的时候？

20 医学新浪潮

与 Michael Le Page 一起探讨针对“不可成药”蛋白质的药物

22 光圈

震撼影像捕捉了渔网留下的破坏

24 本周之书

一种思考生命的新方式

5 火云 危险的风暴正在法国助长火势

特写

30 预测机器

一种新型AI正开始理解物理世界中的因果关系

36 CSI：微生物组

我们能否利用隐藏的微生物线索来破解犯罪？

40 悲伤发作

失去挚爱可能会导致突然的类似恐慌发作的片段

封底页

45 益智游戏

尝试我们的填字游戏、谜题和快速测验

46 几乎是最后的话

家谱在什么时候停止翻倍并开始减少？

47 Tom Gauld 为《新科学家》创作

漫画家眼中的世界

48 反馈

探讨气球动物的几何学

48 Twisteddoodles 为《新科学家》创作

描绘生活轻松的一面

2026年8月8日 / 《新科学家》 / 1

超过 10,000 辆电动汽车，且在持续增加。

今年，我们将欧洲用于配送的电动汽车增加到 10,000 辆以上。 亚马逊

基于对 2026 年底前英国和欧盟境内预计配送服务合作伙伴车辆的估算。

社论

欧洲面临炽热的新现实

矛盾的是，抑制山火也使得欧洲国家在面对巨型火灾时更加脆弱。专家表示，预防策略需要更新，Alec Luhn 报道

随着法国和西班牙持续遭受前所未有的山火袭击，责任追究之争也愈演愈烈。右翼反对派政治人物声称，由于近一半的 Canadair 灭火飞机处于停运状态，两国政府在水上灭火飞机方面的投入不足。官员们则指责由保守派领导的地区（如马德里，其森林防火部门的人员编制仅为一半）破坏了火灾控制工作。

但专家表示，关于灭火的争论关注的是症状而非病根。相反，欧洲必须在防火预防措施上投入更多，例如通过有意识地引发受控火灾来管理植被。这包括英国，在萨福克郡的一场山火距离赛兹韦尔（Sizewell）核电站不足 5 公里，而另一场火灾则烧毁了苏格兰 Cairngorms 25 平方公里的荒原和森林。

“你可以将飞机的数量增加一倍，甚至三倍。这将非常、非常昂贵，而且你投入在那里的每一欧元，就意味着少了一欧元投入到预防中，”法国山火咨询公司 Warucene 的 Sebastien Lahaye 表示。“灌木和丛林大量生长，因此那些一旦失控的火灾更有可能迅速扩大”。“尽管如此，你仍然会遇到一场或几场失控的火灾，并演变成这种巨型火灾。”

一场灾难性山火的要素是干旱、高温和燃料的积累，而今年这三者都分量十足。欧洲部分地区潮湿的春季增加了植被，随后这些植被在今年夏天广泛的干旱中枯萎。现在，欧洲大陆的大部分地区正遭受第四波热浪的袭击。

整个欧盟境内已有超过 4300 平方公里的土地被付之一炬，如果这种情况持续下去，将打破去年 1.08万,800 平方公里的过火面积记录。波尔多和马德里附近创纪录规模的山火迫使当地进行了二战以来最大规模的撤离行动之一。

根据荷兰格罗宁根大学的 Raúl Cordero 及其同事上周发表的一项研究，自 1981 年以来，南欧每年山火的平均数量和过火面积实际上有所下降。地中海国家警告公众不要在林区生营火或抽烟，并通过培训消防员以及购买飞机和直升机，提高了快速扑灭火灾的能力。

但这种严厉的压制反而制造了一个“火药桶”。灌木和丛林大量生长，因此那些一旦失控的火灾更有可能演变成极端的巨型火灾。这种“抑制悖论”因人们从农村向城市的迁移而进一步加剧，导致以前的农场和牧场杂草丛生。与此同时，在巴塞罗那等城市周边的森林中，更多的房屋拔地而起，增加了山火的风险，而山火大多是由人类点燃的。

现在，全球变暖正在抵消山火抑制能力的提升。研究发现，在南欧大部分地区，炎热、干燥且多风的“火灾天气”天数增加了一倍多。

“今年发生的情况似乎证实了我们的假设……即我们即将进入一个大火频发的新时代，而这个时代此前在欧洲并不存在，是因为成功的火灾抑制和教育，”Cordero 表示。

消防员在阿维拉（Ávila）与大火搏斗，这场火灾是西班牙历史上规模最大的一次

欧洲不能仅以加强压制来应对，而必须承担起使景观具备防火韧性这一往往令人不安的任务。德国马克斯·普朗克化学研究所的约翰·戈尔达默（Johann Goldammer）表示，粗略来看，许多国家的防火预算中，约 90% 用于灭火，而只有 1万% 用于预防。他说：“情况应该是相反的。”欧盟一直致力于到 2030 年在其 20% 的陆地面积上恢复自然，以应对气候和生物多样性危机。但戈尔达默认为，在许多情况下，林地需要更密集地管理，包括通过可持续伐木。

仅靠伐木是不够的。波尔多附近的火灾在朗德（Landes）席卷了欧洲最大的人工林，这片 10,000 平方公里的土地曾经如此沼泽化，以至于牧羊人必须踩着高跷行走。拿破仑三世通过种植海滨松 reclaim 了这些沼泽，这些松树至今仍被用于生产纸张和木材。特别是在 2022 年那里发生的一场毁灭性火灾之后，土地所有者修建了进入道路、排水沟和防火隔离带。但他们同时也重新种植了单一的松树林，这种做法增加了火灾风险。葡萄牙类似的桉树单一林也助长了像佩德罗冈格兰德（Pedrógão Grande）火灾那样的烈火，该火灾在 2017 年造成 66 人死亡。拉海（Lahaye）表示，虽然在朗德实现树种多样化可能很困难，但至少需要拓宽并增加防火隔离带。

以火攻火

即使在森林中没有清理至裸地的防火带，也可以通过疏伐树木和清理灌木来创建燃料隔离带。重新建立果园、葡萄园、牧场和农业有助于创建 4 / New Scientist / 8 August 2026 ---

“火云”正使山火变得更加危险

……一个由高低不同火燃料载量组成的“马赛克”格局，而非像波尔多和马德里火灾那样，由于连续不断的树冠覆盖而导致火势失控地爆发。西班牙部分地区已部署羊群、绵羊群和牛群来啃食易燃的灌木。但在法国，由于山火监管条例与生物多样性监管条例之间的冲突，仅有 30% 的危险灌木被清理。清理林下植被最有效的方法通常是通过计划烧除。2020 年美国西部的山火在此前进行过计划烧除的区域，其严重程度降低了 16%，且排放的有害颗粒污染物也减少了 14%。但欧洲的人口密度要高得多，当地人通常将这种技术视为一种威胁。在某些地区，法规仍然经常禁止这种做法。虽然在朗德省的道路沿线开展了计划烧除，但木材种植园则被排除在外。■ 积雨云（PYROCUMULONIMBUS cloud），一种由极高温山火产生的雷暴云，首次在法国被雷达观测到。这些由上升的热空气、烟雾和水分组成的致命“火云”会通过强风和飞溅的余烬加速火势，预计在欧洲将变得更加普遍。“你会看到一个倾斜的柱状体在吸入空气，然后将余烬向前传播，同时它还会产生闪电。这是点燃火线前方燃料的另一种方式，因此这些电暴单元是不可阻挡的，”法国科西嘉大学的 Jean-Baptiste Filippi 表示。据地区官员称，在法国吉伦特省靠近波尔多市（见左侧主报道）的山火中，分别在 7 月 24 日和 7 月 25 日产生了积雨云。余烬和闪电在原有的炼狱周围点燃了新火源。积雨云有时被称为“喷火的云龙”，在法国索莫斯村附近观测到的一朵火云，通常由北美、西伯利亚和澳大利亚森林中的大规模山火产生。它们催生了诸如 2009 年在澳大利亚造成 173 人死亡的“黑色星期六”丛林大火等臭名昭著的火灾。然而在欧洲，仅在西班牙和葡萄牙确认过少数几起积雨云案例，包括“数十次闪电击中可点燃远在主火线前方 100 公里处的火灾”，其中一次发生在 2017 年葡萄牙毁灭性的佩德罗冈格兰德大火期间，该火灾摧毁了数百栋房屋并造成 66 人死亡。但这种情况似乎正在改变。“你必须同时具备两个条件：一个极端的发热源和一个与火相互作用的不稳定大气，”欧洲中期天气预报中心的 Francesca Di Giuseppe 表示。“如果……火灾产生的热量增加，就像我们的研究预测的那样，那么是的，你应该会看到更多此类事件。”当高达 800℃ 的温度将烟柱送入天空时，这场火风暴开始成形。随着热空气上升并冷却，水蒸气在灰烬上凝结形成火积云。如果烟柱足够热且大气足够不稳定，云层可以继续上升，达到 1万,000 米的对流层顶，这正是客机的巡航高度。它有时会到达平流层，散播相当于一座中型火山的烟雾量。当冰晶碰撞产生的静电引发雷电时，它就变成了积雨云。尽管云层降下的雨水通常在热空气中蒸发，但这种雷暴依然能创造出自己的天气系统。据 Filippi 称，云层的上升气流会吸入空气，产生高达 160 公里的每小时风速来助燃，并将火势蔓延速度加快至多五倍。

与此同时，强烈的下沉气流会产生不稳定的阵风，将余烬带走并引发新的火灾。数十次闪电击中可点燃远在主火区前方 100 公里处的火势。它还可能产生由扭曲火焰组成的火龙卷。随着风力、余烬和闪电增强并扩大野火规模，它会释放出更多的热量和烟雾，这种反馈循环可以让上升的云团在数小时内持续增强。大气高层的风切变和地面的不稳定阵风都可能使火势向不可预测的方向蔓延，使得遏制其加速或规划有针对性的疏散工作几乎成为不可能。■ AL

© 2017 SHANDOLPH / PA / SHUTTERSTOCK 2026年8月8日 / New Scientist / 5 --- 新闻 神经科学

痴呆症手术引发质疑

据 Carissa Wong 报道，中国有数千人接受了一项现已被禁止的实验性手术，该手术旨在帮助大脑清除与阿尔茨海默病相关的蛋白质。

据信，中国有 2000 多人接受了治疗阿尔茨海默病的实验性手术，据称该手术曾帮助一名卧床不起的患者重新行走。这项手术通过将淋巴系统与附近的静脉连接以清除阿尔茨海默病相关蛋白质，目前在该国已被禁止。

“当我第一次听说这种疗法时，我感到很震惊，”瑞士伯尔尼大学的 Steven Proulx 表示，“基本上没有任何证据支持这一点。”

但本月早些时候在伦敦举行的阿尔茨海默病协会国际会议上展示的美国和新加坡小型安全性试验的初步结果表明，该方法在最多六个月内是安全的，并出现了一些认知改善的迹象。

目前一项正在进行的随机试验正在将该手术与伪手术（sham procedure）进行比较，这应该能更清晰地揭示其潜力，但有些人对该试验的伦理问题及其有效性提出了质疑。

这种被称为深颈淋巴静脉吻合术（DCLVA）的手术，旨在增强大脑废物的淋巴系统排泄。淋巴系统是一个由血管和淋巴结组成的网络，负责收集组织中的废物以便分解或排出体外。其核心理念是通过手术将该网络连接到附近的静脉，从而加速这一过程，更高效地清除 beta-淀粉样蛋白和 tau 蛋白等阿尔茨海默病相关蛋白。

一张显示阿尔茨海默病女性大脑的计算机断层扫描（CT）图像

该技术由中国杭州求是医院的谢庆平开发，基于一种治疗慢性肢体水肿的常规疗法。2024 年，谢庆平及其同事在《整形与重建外科》（Plastic and Reconstructive Surgery）杂志上报道了他们对 50 名阿尔茨海默病患者进行了 DCLVA 手术。在发稿前，谢庆平未回应《新科学家》的置评请求。

“目前尚不清楚任何改善是否足以让患者察觉”

他们还发布了一段视频，显示一名此前卧床不起的阿尔茨海默病患者在手术八个月后快步行走。但一些人对该病例报告的可靠性表示担忧。

“一个患者的视频我不会认为它是可信的，”Proulx 表示，他补充说研究人员没有评估其他改善情况，例如该患者脑成像扫描结果的变化。

不过，佐治亚州奥古斯塔大学的 Fan Fan 表示，这段视频在中国走红。到 2025 年 5 月，据估计中国约 300 家私立和公立医院有数千人接受了该手术。

Proulx 表示，报告显示其中许多手术是由整形外科医生完成的，费用从数万到数十万人民币（数千到数万英镑）不等。

缺乏证据

但在 2025 年 6 月，中国国家卫生健康委员会禁止在研究环境之外实施该手术，理由是缺乏安全性和有效性的证据。在发稿前，该委员会未回应《新科学家》杂志的置评请求。

“中国停止将 DCLVA 用于治疗阿尔茨海默病是恰当的，”曾在哈佛大学研究阿尔茨海默病的顾世久（Shiju Gu）表示。

迄今为止，约有十余篇病例报告和临床试验报道了对中度或重度阿尔茨海默病患者实施 DCLVA。在近期的一次阿尔茨海默病会议上展示的美国和新加坡临床试验表明，该手术在短期内是安全的。新加坡的试验还发现，四名参与者中有两名出现了轻微的认知改善。

西雅图华盛顿大学的杰夫·伊利夫（Jeff Iliff）表示，这些改善不一定是微不足道的，但这些试验没有一项是随机或双盲的。

“基于目前可获得证据，我会将 DCLVA 的有效性描述为不确定，因为中国现有的临床研究通常是非对照的、短期的且在方法论上存在局限，这使得很难将真实的手术效果与安慰剂效应、术后护理或患者病情的波动区分开来，”顾世久说道。

挪威奥斯陆大学的佩尔·克里斯蒂安·艾德（Per Kristian Eide）表示，目前尚不清楚任何改善是否足以让患者或其家属察觉。

一项将 DCLVA 与伪手术进行比较的随机对照试验目前正在进行中。该试验由中国沈阳军区总医院的陈辉胜（Hui-Sheng Chen）领导，在发稿前他未回应《新科学家》杂志的置评请求。该试验旨在招募 50 名轻度至中度阿尔茨海默病患者。结果预计将于明年公布。

6 / 《新科学家》 / 2026 年 8 月 8 日 --- 人类进化

我们基因组中都有来自“幽灵”祖先的 DNA

迈克尔·勒佩奇（Michael Le Page）

伦理担忧

尽管之前的试验因缺乏对照组而受到批评，但仍有人对伪手术的伦理问题提出质疑。“对于老年且可能体弱的人群来说，任何形式的手术及其随之而来的麻醉所带来的风险，使得执行一次‘虚假’干预很难证明其合理性，尽管这对整体研究至关重要，”伊利夫（Iliff）说道。同样持有这些伦理担忧的艾德（Eide）还指出，该试验的方案称将在手术四天后观察认知能力的改善。“你不会期望在这么短的时间内看到任何改善，”他说。然而，这种手术可能有一个重大优势：它是一次性的干预。参与新加坡试验的樟宜综合医院（Changi General Hospital）的文森特·泰（Vincent Tay）表示，这使得 DCLVA 值得追求。阿尔茨海默病的治疗通常侧重于缓解症状，但像lecanemab这样能清除大脑中β-淀粉样蛋白团块的抗体疗法已在美国和英国获得批准。但由于其疗效存疑且成本高昂（部分原因是需要通过输液给药），这意味着英国的医疗服务无法提供此类治疗。“优势在于这[手术]是一次性的，”加利福尼亚州斯坦福大学（Stanford University）的拉格·艾兰（Raag Airan）说道。那么，这种手术有一天会在中国甚至全世界成为常规操作吗？“我正在密切关注该领域公布的结果，所有研究这个问题的专家也一样，”伊利夫说道。■

在现今的所有活人中，有 0.5 到 1% 的 DNA 来自一个未知的古人类物种，该物种在智人（Homo sapiens）离开非洲之前曾与他们发生过杂交。不同的人拥有来自这个“幽灵”祖先的不同 DNA 片段，而且至关重要的是，这些片段相当均匀地分布在整个基因组中。这意味着，如果能将所有片段拼凑在一起，或许有可能组装出这个古老物种几乎完整的基因组。

早期的研究利用化石中的古 DNA 序列揭示，智人在过去 5万,000 年中与尼安德特人和丹尼索瓦人发生过杂交。虽然之前的研究也识别出至少来自另外两个谱系的幽灵祖先，但加州大学伯克利分校（University of California, Berkeley）的普里亚·穆尔贾尼（Priya Moorjani）及其同事使用了一种功能强大得多的新方法。该方法基于为基因组的部分区域构建“家族树”，而这在计算能力上直到最近才成为可能。“我相信这是人类遗传学研究的下一个前沿，因为它让我们甚至能够了解那些没有任何化石 [DNA] 证据的祖先，”穆尔贾尼说道。

尼安德特人和丹尼索瓦人的基因组中只有很小一部分共同保留在我们体内。“基因组的长片段要么完全没有，要么只有极低比例的尼安德特人或丹尼索瓦人血统，”团队成员、马里兰州约翰斯·霍普金斯大学（Johns Hopkins University）的阿俊·比丹达（Arjun Biddanda）说道。“然而，我们能够在这些区域中发现较高比例的这种幽灵血统。”

研究人员认为，这是因为尼安德特人和丹尼索瓦人的种群规模非常小，因此更有可能出现有害突变。“在这些区域，很可能存在针对尼安德特人或丹尼索瓦人血统的自然选择，”穆尔贾尼说道。但来自一个幽灵祖先的更多 DNA 似乎被选择了，或者至少在进化意义上是中性的。她表示，这可能是因为这个幽灵祖先的种群规模要大得多。该团队的工作表明，这个幽灵祖先在大约 80万,000 年前从导致现代人类的主要谱系中分离出去。随后，这两个谱系在一段时间后再次汇合并发生了杂交（Science, doi.org / rhtb）。

“由于所有现代人类都携带这种幽灵祖先的特征，因此基因流最有可能发生在非洲，且该种群存在于非洲，”穆尔贾尼（Moorjani）说道。研究人员推测，基于所有这些因素，这个幽灵祖先可能是一个已通过化石为人所知的物种，例如海德堡人（Homo heidelbergensis）。通过他们的方法识别出的第二个幽灵祖先则古老得多。它在大约 180 万 年前从导致现代人类的主谱系中分离出来，但后来在欧亚大陆与丹尼索瓦人（Denisovans）发生了杂交。穆尔贾尼表示，现代人类通过丹尼索瓦人继承了它的 DNA。伦敦自然历史博物馆的克里斯·斯特林格（Chris Stringer）认为，这个更古老的幽灵谱系最可能的候选者必须是直立人（Homo erectus）。“这些结果令人兴奋，”斯特林格说，“它们提供了间接的方法，用以重建像直立人和海德堡人这样更原始物种的部分基因组。” ■ 8 August 2026 / New Scientist / 7 --- 新闻 考古学

亚马逊巨大的失落文明

詹姆斯·伍德福德（James Woodford）发现，鲜为人知的 Aquiry 社会在雨林中建造了数万座巨大的土方工程，人口最高可达 300 万。

使用能够穿透林冠的激光对茂密的亚马逊雨林进行测绘，揭示了一个繁荣近 1500 年的失落文明所建造的数百座庞大结构。这些发现为之前的观点提供了有力支持，即如今被视为荒野的地区曾居住过数百万人。

Aquiry 文明大约存在于公元前 600 年至公元 850, 年，在不使用金属工具或动物的情况下，建造了巨大的几何形土方工程，其形状可能代表各种神灵。其中许多工程还由古道连接。

芬兰赫尔辛基大学的马蒂·帕西宁（Martti Pärssinen）表示，直到近几十年，科学界才开始了解这些人，因为他们居住的亚马逊地区曾被认为是“几乎无人居住的原始雨林”。

但研究人员一直在利用激光雷达（lidar）进一步了解该文明的规模。这是一种激光扫描技术，可以创建大多数植被下方地面的三维地图。从飞越森林的飞机上，每平方米向地面发射约 3万 次激光脉冲。林冠中的微小缝隙允许部分脉冲到达下方地形并返回飞机。

2024, 帕西宁及其同事在巴西的阿克里州（Acre）和亚马逊州（Amazonas）近 4500 平方公里的区域上空进行了飞行。测绘结果揭示了 432 个被称为“地画”（geoglyphs）的带沟几何围合结构，其中近 400 个是科学界首次发现。

通过将这些结果推算到 Aquiry 文明居住的 18.3万,000 平方公里领土上，研究团队估计可能存在多达 30,000 个地画。

平均而言，这些土方工程覆盖 3 公顷，但最大可达 50 公顷。沟渠通常宽约 12 米，深达 5 米。

“这些建筑的规模和精度简直令人震惊，”帕西宁说。研究人员估计，每个平均规模的地画大约需要 300 人来建造和维护。

帕西宁表示，研究结果表明，在该文明于公元 100 年至 300, 年间的巅峰时期，总人口在 125 万 到 300 万 之间，且他认为上限估计的可能性最大。

他说：“亚马逊是未经触碰的荒野这一观念在很大程度上是一个神话，[事实是] 古代人类社会在塑造如今的热带雨林中发挥了重要作用。”

在过去的 20 年里，帕西宁及其同事在巴西的阿克里州、亚马逊州和朗多尼亚州（另一个巴西州）以及玻利维亚挖掘了 40 多个祭祀和定居点遗址。

令人震撼的规模

他说，这表明这些地画是仪式中心，被用作“集会、政治决策、展示首领慷慨之举、仪式舞蹈、音乐、宴饮、体育竞赛和球类游戏的场所”。容纳数个家庭的长形住所可能以分散的形式分布在土方工程周围（Nature, DOI: 10.1038 / s41561-026-02054-6）。Pärssinen表示，Aquiry对亚马逊雨林产生了重大影响，他们焚烧竹林以种植玉米、南瓜和木薯，并为了建造巨大的土方工程及其延伸的广泛道路而砍伐森林。Aquiry还种植了许多结果树种，包括巴西栗树和桃形棕榈。目前尚不清楚导致他们灭绝的原因。

佛罗里达大学的Michael Heckenberger表示，关于人口规模的说法虽然大胆，但“具有合理性且极其重要，甚至令人震撼”。他说：“在过去二十年里，关于哥伦布时代前亚马逊南部城市化的观点已获得关注。我们现在需要了解的是，在可探测到的土方工程周围发生了什么，例如能够支撑如此高估计人口的居住地。”

巴西马瑙斯亚马逊国家研究所的Charles Clement表示，这项新研究“在亚马逊森林中确立了一种大规模文化”。他说：“毋庸置疑，他们还没有对所有新发现的地画进行实地验证，但之前的激光雷达研究效率极高，因此我觉得他们的数字是可靠的。” ■ 8 / New Scientist / 2026年8月8日

---

天文学

捕捉到一颗异常的游荡超大质量黑洞

Leah Crane

一个所谓的“游荡超大质量黑洞”最极端的例子刚刚被发现。黑洞不发射光，因此马里兰大学的Robert Stein及其团队只有在它吞噬一颗恒星（即所谓的潮汐瓦解事件，TDE）时才发现了它。一种机器学习算法在加利福尼亚州兹维基瞬态设施（Zwicky Transient Facility）的数据中将这次光暴识别为可能的TDE，随后研究人员利用其他几台望远镜进行了后续观测以予以确认。

这次TDE是我们见过的距离其宿主星系中心最远的一次。它距离其星系中心超过3万,000光年——作为对比，我们的太阳系距离银河系中心约2.6万,000光年（The Astrophysical Journal Letters, doi.org / hccvkm）。

游荡的超大质量黑洞通常是因为合并而开始远离星系中心的旅程：两个星系猛烈碰撞，其中一个中心黑洞被抛出。寻找这些黑洞可以告诉我们关于其星系历史和增长的信息。

“这个结果令人兴奋的地方在于，我们认为这只是冰山一角，”Stein说。“如果我们能了解有多少个游荡的超大质量黑洞，那么它就能告诉我们该星系经历了多少次合并，以及它是如何随时间演化的。”

关于这个特定黑洞及其周围环境，我们能了解的信息是有上限的。被吞噬恒星发出的光仍然太亮，导致无法看清周围的其他任何东西。可能有一个与该黑洞在引力上绑定的小型星团，其中一颗恒星在坠入时引发了TDE，但在光芒消退之前，我们无法确认这一点。 ■ 健康

减少蛋白质摄入能否延长我们的寿命？

Carissa Wong

从谷物到薯片再到咖啡，富含蛋白质的产品充斥在超市货架上，但越来越多针对非人类动物的证据表明，限制蛋白质摄入可能会减缓衰老并延长寿命。

威斯康星大学麦迪逊分校的 Dudley Lamming 表示，根据最新的证据综述，“对于久坐不动的人或动物，增加超出需求的饮食蛋白质会导致代谢问题，并可能缩短寿命”。

Lamming 和他的同事，同样来自威斯康星大学麦迪逊分校的 Bailey Knopf，筛选了 300 多项研究（主要针对久坐不动的小鼠和大鼠），探讨将蛋白质摄入量限制在防止缺乏所需的最低限度如何影响衰老和寿命。

在这些研究中，啮齿动物通常要么摄入相当于美国指南建议的最低蛋白质需求量（每公斤体重 0.8 克蛋白质），要么摄入大约两倍的量。后者与美国饮食指南近期为实现最佳健康而建议的摄入量相似，即每公斤体重 1.2 到 1.6 克蛋白质。

在一项研究中，摄入量接近最低蛋白质需求的大鼠寿命延长了近 120 天，比摄入量约为两倍的大鼠长了 50% 以上（Cell Press Blue, doi.org / rh2n）。

Lamming 表示，这可能是因为限制蛋白质会提高激素 FGF21 的水平。在啮齿动物中，这种激素作用于大脑、脂肪组织和肝脏，以改善血糖水平并减少衰老细胞的数量。这些细胞随年龄增长而积累，它们已停止分裂并会导致慢性炎症。

研究还表明，限制蛋白质摄入可以抑制小鼠体内一种名为 mTORC1 的蛋白质的活性。mTORC1 通常刺激细胞生长，但降低其活性可增强细胞修复。

Lamming 说：“它基本上将细胞从增殖状态（即生长并进行大量代谢的状态）切换到分裂更慢（如果还分裂的话）并专注于修复的状态。”他表示，研究表明这降低了小鼠患年龄相关疾病的风险并延长了寿命。

所有这些迹象表明，近期增加蛋白质摄入的趋势对于经常锻炼的人来说可能是没问题的，因为他们可能需要更多蛋白质来修复肌肉损伤。但 Lamming 表示，对于久坐不动的人来说，摄入超过需求量可能会造成伤害。

“对于久坐不动的人，增加超出需求的饮食蛋白质会导致代谢问题”

但伊利诺伊大学厄巴纳-香槟分校的 Donald Layman 指出，在蛋白质代谢方面，啮齿动物与人类存在巨大差异。他表示，因此限制蛋白质可能会以不同的方式影响不同物种的 mTORC1 通路。

此外，他表示，该综述并未包含大多数关于人类的研究，而这些研究表明，将蛋白质摄入量提高到最低限度以上可以降低我们衰弱的风险。

Lamming 表示，没有研究被刻意省略。他认为，除了增强肌肉和预防衰弱之外，关于高蛋白质摄入是否有助于抗衰老的证据尚不统一。

例如，一项研究将此与英国老年双胞胎中年龄相关性肌肉流失（即肌少症）风险的降低联系起来。但英国谢菲尔德大学的 Elizabeth Williams 表示，肌少症的发病率如此之低，以至于这种联系并不可靠。■

2026年8月8日 / New Scientist / 9

新闻 技术

量子优势倍增

Karmela Padavic-Callaghan 表示，量子计算机已经解决了三个传统计算机无法破解的问题，展现了它们在科学领域潜在的重要性。

量子计算机正进一步深入“不可能”的世界。三个关于无法由其他任何类型机器完成的计算新案例表明，此类计算机正成为有用的科学工具。

在科学研究中，许多情况需要进行大量的计算机计算才能解决问题，例如尝试预测大分子的行为。但在某些情况下，计算过于复杂，传统计算机永远无法得出答案。这就是量子计算机可以产生“量子优势”的时刻，因为它们可以利用量子现象来产生解决方案。

但如果一个答案只能由一台机器产生，你如何知道是否可以信任它？IBM 的研究人员及其同事现在声称发现了三个量子优势的新案例，并就为什么这些计算值得信赖提出了论据。

其中两个案例是对数学模型的模拟，这些模型对于研究材料的物理学家具有特殊意义。在其中一个案例中，该模型旨在捕捉量子材料在被激光照射时的行为，重点研究一种磁体。利用量子计算机，研究人员发现该磁体的一些属性会发生振荡，而这在以前是不为人知的。值得注意的是，在日本的“富岳”（Fugaku）超级计算机以及 Nvidia GPU 上的模拟均未能产生结果。

在第二个案例中，团队重点研究信息IBM 研究人员已利用模拟来测试量子机器在一种极不均匀的材料中是如何被扰乱的，这可以帮助研究人员更好地理解在例如使用催化剂加速化学反应时所发生的混沌过程。在此，研究人员将量子计算机的输出与传统计算机的一种领先模拟方法进行了比较——而该方法在过去曾多次驳斥量子优势的主张。

“你实际上可以从实验中获取数据，并估算出运行效果如何”

对于某些模拟参数，这些方法无法与量子计算机相媲美。对于这两项模拟，团队通过在几台不同的量子计算机上重复模拟，来努力估算量子计算机的可信程度，每台计算机的噪声量略有不同，而且他们还故意在某些计算运行中注入噪声。IBM 的 Abhinav Kandala 表示，通过对这些噪声程度不同的计算进行比较，团队能够量化量子计算机的输出中有多少是可信的，多少仅仅是噪声。

第三个案例聚焦于一个采样问题，该问题研究具有预设属性（例如内部数学运算的分布）的许多量子计算电路的输出。研究人员已经知道采样问题对于传统计算机来说很困难，但尽管量子计算机可以输出答案，它此前一直难以验证该答案的意义——即它是否陷入了错误之中。团队设计了一套程序，其中涉及对计算进行微调然后重复，以量化量子计算机答案的保真度。

Kandala 表示，量子计算机现在的运行方式正是我们一直希望的样子。“我希望人们开始将它们作为具有可信结果的科学仪器来使用。”

构建工具箱

这些旨在验证量子计算机运行效果的新尝试是朝着正确方向迈出的一步，瑞士苏黎世联邦理工学院（ETH Zurich）的多米尼克·汉格莱特（Dominik Hangleiter）表示，他并非该团队成员。“你实际上可以利用实验数据，对实验运行的质量进行评估。”他表示，这一过程仍不理想，因为它并非完全独立于量子计算机的特性，但对于我们目前拥有的规模仍然较小且噪声较大的量子计算机来说，这是必要的。

除了在验证方面进一步改进，材料的量子计算机模拟也需要进一步推进。以色列巴伊兰大学（Bar-Ilan University）的埃马努埃莱·达拉·托雷（Emanuele Dalla Torre）表示，磁体模拟从基础科学的角度来看取得了重大发现，但与现实世界的磁体相比，它仍然过于理想化。“问题本身归根结底是一个基础科学问题，目前[不一定]有明确的用途。”

但该团队对这些新进展如何开启量子实用化时代持乐观态度。IBM的杰伊·甘贝塔（Jay Gambetta）表示：“我们正在构建一套工具箱，用于确定我们如何信任[量子计算机的]输出。一旦我们能够信任输出，就可以开始确定具体应用了。” ■

COWNE JINGUIN 10 / New Scientist / 8 August 2026 --- Archaeology

400年前智利木乃伊中的天花病毒源自欧洲

詹姆斯·伍德福德（James Woodford）

从印加木乃伊中采集的DNA样本证实，其中存在一种最早在欧洲出现的天花病毒株。爱尔兰都柏林三一学院（Trinity College Dublin）的重树·中込（Shigeki Nakagome）表示：“我们将此解读为天花通过欧洲殖民进入美洲的第一个直接分子证据。”

几个世纪以来，天花一直是人类的灾难，直到20世纪80年代通过疫苗接种在除少数高安全性医学实验室之外的所有地方被根除。该病毒的严重毒株死亡率高达三分之一，幸存者身上会留下由覆盖皮肤且充满脓液的水疱和结痂所形成的疤痕。在被欧洲人殖民的世界部分地区（如澳大利亚和美洲），由于原住民没有天然免疫力，这种疾病尤为残酷，导致数百万人死亡。

在美洲，历史记录显示，在欧洲殖民者抵达后的16世纪初，天花开始摧毁原住民社区。英国邓迪大学的马修·沃德（Matthew Ward）并未参与这项新研究，他表示，包括天花在内在美洲传播的几波流行病是历史上最重要的事件之一。他说：“这些疾病可能杀死了美洲原住民人口的95% 弗朗西斯科·皮萨罗（Francisco Pizarro）领导了西班牙对印加帝国的征服

FOTOCA B. ARSKETT MAGS 并为欧洲殖民铺平了道路。”

中込及其同事研究了在智利北部一个名为Camarones 9的考古遗址发现的木乃伊骨骼样本，该遗址于1957年被发现。在随后的几十年里，该遗址共出土了41具包裹在毯子中并随葬陪葬品的木乃伊。中込说：“虽然最初的目标是研究人类DNA，但我们也搜寻了古代病原体的DNA。就在那时，我们在两个个体中发现了与天花病毒匹配的片段。”

第一个个体是死时年龄在30到35岁之间的女性；第二个是18到20岁的男性，碳定年法显示其死亡时间在1492和1631年之间。通过重建病毒基因组并将其与其他古代和现代天花病毒进行比较，团队发现它们属于一个天花谱系，其中包括中世纪早期和近代早期的欧洲毒株（《科学》杂志，doi.org / rhsr）。

这一发现增加了这样一种可能性：木乃伊躯干上的微小皮肤病变（此前被归因于长期接触砷）实际上可能是天花患者皮肤上特有的脓疱疤痕。团队成员、圣地亚哥智利大学的康斯坦莎·德拉富恩特·卡斯特罗（Constanza de la Fuente Castro）表示：“这一发现使我们能够识别出一小群受到病毒影响但没有历史记录的人。”她说，在这次新研究之前，人们仅通过大规模爆发的历史记载才知晓这种病原体。

这项工作由于对原始遗骸的干扰以及分析过程中材料的损失而显得尤为敏感，这两点在原住民和科学界都被认为是极具争议的话题。德拉富恩特·卡斯特罗说：“我们永远不会忘记，每一次DNA提取的背后是一个人，而不是一个数据点。” ■ 健康

吉兰-巴雷综合征药物可关闭部分免疫系统

首个针对吉兰-巴雷综合征（Guillain-Barré syndrome）的靶向治疗药物在后期临床试验中取得了良好效果。“在我们的研究中，患者在各项指标上的康复速度更快，且康复得更彻底，”开发该疗法的加州公司 Annexon Biosciences 的 Doug Love 表示。

吉兰-巴雷综合征是一种罕见疾病，任何近期有过感染的人都可能患病。在对抗感染后，一个人的免疫系统会开始错误地攻击其神经。这种攻击会导致手臂和腿部出现无力或刺痛感，随后可能进展为行走、说话、吞咽和呼吸困难，有时会导致瘫痪或死亡。

这种名为 tanruprubart 的新疗法是一种单克隆抗体，其作用机制是阻断血液中一种名为 C1q 的蛋白质。这会关闭免疫系统的一个分支，即经典补体途径（classical complement pathway），而该途径在吉兰-巴雷综合征中被误导去攻击神经。

在孟加拉国和菲律宾进行的一项后期试验中，241 名近期被诊断患有该病的人员接受了 tanruprubart 静脉输注或安慰剂输注。平均而言，接受药物治疗的人员减少了 28 天的机械通气时间和 7 天的重症监护时间，并且能够提前 31 天独立行走。与接受安慰剂的人员相比，接受 tanruprubart 的参与者在两个月后获得了更高的“平均而言，接受药物治疗的人员减少了 28 天的机械通气时间”生活质量和行动能力。

这些结果在 6 月于瑞士日内瓦举行的欧洲神经病学学会（European Academy of Neurology）大会上公布。试验中未出现严重副作用，且尽管 tanruprubart 关闭了部分免疫系统，但它似乎并未增加参与者感染的风险。

基于这些结果，Annexon Biosciences 最近已向欧洲药品管理局（European Medicines Agency）提交了 tanruprubart 的审批申请。■

Alice Klein 2026年8月8日 / New Scientist / 11

---

新闻 健康

眼睛植入物部分恢复视力

一款让严重视力受损者能够阅读的设备已获得欧盟和英国的批准 Oluoma Aneke

一种能够改善常见失明形式患者视力的视网膜植入物已获准在欧洲使用。这是此类设备——能够让某种严重视力受损的人阅读——在全球范围内首次获得批准。

“我们有患者能够通过该设备认出女王的面孔，还有人画出了悉尼歌剧院，”位于加利福尼亚州阿拉米达的 Science Corporation 公司的 Darius Shahida 说道，该公司是该植入物的研发方。

这款名为 PRIMA 的植入物能为地理萎缩症患者恢复功能性中心视力——即观察精细细节和阅读所需的正前方视力。地理萎缩症是一种进行性且不可逆的干性年龄相关性黄斑变性，其特点是随着视网膜细胞的死亡且无法再生，导致视网膜中被称为黄斑的部分缓慢恶化。

目前针对地理萎缩症的治疗手段极其有限。在美国，有两种通过眼部注射给药的药物获得批准，用于阻断可能损害视网膜的蛋白质。

该植入物——其大小约为针头大小并被植入视网膜下方——与一副带有框架安装摄像头的眼镜配合使用。摄像头捕捉并处理图像，然后使用近红外光将结果投射到植入物的芯片上。我们无法看到这个波长的光，因此该过程不会干扰任何现有的视力。随后，这些芯片将近红外光转换为电脉冲，以刺激视网膜中被称为双极细胞的神经元，这些细胞通常不会受到黄斑变性的损害。

通常情况下，双极细胞与视杆细胞和视锥细胞协同工作，后者将光线转换为电信号，且会受到黄斑变性的影响。PRIMA 系统绕过了视杆细胞和视锥细胞，使双极细胞能够产生可被大脑解读的电信号。

该眼镜直接从光线中产生电能，类似于太阳能电池板。“这意味着患者不需要连接电源，”Shahida 说道。

在去年 10 月发表的一项小型研究中，PRIMA 在约 80% 的病例中恢复了中心视力并提高了视敏度。

“美国的正常视力是 20 / 20，而失明是 20 / 200，”同样来自 Science 公司的 Frank Brodie 说道，“使用 PRIMA 的患者达到了约 20 / 160，所以他们从失明状态变成了非失明状态。”

大多数副作用与手术相关，并在两个月内消失。“身体恢复很快，有些人一天之内就感觉好多了，”Bodie 说道。

宾夕法尼亚州费城威尔斯眼科医院（Wills Eye Hospital）的 Sunir Garg 表示，在这一干预措施被广泛使用之前，还需要在更大规模的人群中进行更多研究，以获取长期安全性数据。他表示，研究人员还需要揭示该植入物是否仅对具有某些患者相关或病情相关特征的人群有效。

该干预措施已根据欧盟医疗器械条例获得 CE 认证标志，使其能够在欧盟和英国使用。Shahida 表示，PRIMA 还获得了美国食品药品监督管理局（FDA）的“突破性设备”和“人道主义使用设备”认定，“这是通往美国获批的重要途径”。

■ 古生物学

一头鱼龙有着极其出人意料的最后一餐

在澳大利亚内陆发现的一块化石提供了首个证据，证明大型海洋爬行动物——鱼龙（ichthyosaurs）曾捕食翼龙（pterosaurs），而翼龙是已知最早进化出动力飞行的脊椎动物。

2019年，业余化石猎人在昆士兰州内陆一个半干旱地区、靠近里士满镇（Richmond）的指定公共地点发现了部分鱼龙的化石骨骼。在早白垩世（1.2亿至9500万年前），该地区曾是埃罗曼加海（Eromanga Sea）的一部分，这是一片覆盖了澳大利亚大陆三分之一以上的浅内陆海。

三年后，古生物学家在挖掘该遗址时，发现了一头巨大的、部分关节连接的鱼龙。该生物被鉴定为Platypterygius australis，体长近7米，与成年虎鲸相当。

“令我吃惊的不不仅是动物的体型，还有其骨骼受损的方式，”领导此次调查的澳大利亚新英格兰大学的马特·怀特（Matt White）说道。他解释说，脊椎被劈开，肋骨缺失，且许多骨骼——鱼龙在被吞噬前吃掉了翼龙——显示出巨大的粉碎性咬痕。怀特表示，团队得出的唯一解释是，一个强大的捕食者肢解并咀嚼了这头鱼龙，最有可能的是巨型 the pliosaur Kronosaurus queenslandicus。

团队还在鱼龙的颅骨附近发现了球状肿块。通过对化石周围岩石的机械清除和30次成像分析显示，其中一个球体包含鱼类、类乌贼头足类动物的碎片，以及令人惊叹的两块翼龙骨骼（包括一块下颌碎片）。研究人员认为这是鱼龙的肠道内容物，可能由于消化道破裂而化石化在颅骨附近（发表于《冈瓦纳研究》Gondwana Research，doi.org / hccx5）。

“这几乎肯定是一起捕食或食腐案例，也是首个翼龙被鱼龙捕食的实例，”英国莱斯特大学的大卫·安温（David Unwin）表示。■

詹姆斯·厄夸特（James Urquhart） 12 / New Scientist / 2026年8月8日 --- 分析环境

区域制冷是解决城市极端高温的方案吗？输送冷却水的管道比空调更高效。Alec Luhn 探讨这些管道是否能帮助缓解美国和欧洲遭受热浪袭击的城市。

在欧洲，有超过 7700 万 人享受区域供暖，即由中央工厂通过管道将热水输送到家庭和建筑中。美国拥有数百个此类网络，主要分布在机场、大学和商业区等多建筑园区内。但随着欧洲和北美在一次次热浪中煎熬，许多人开始思考，其镜像版本——冷却水管道网络，是否会比空调更好的制冷选择。

“区域制冷系统的效率是典型建筑内冷水机组的两倍，或者说运行费用仅为后者的一半，”国际区域能源协会（International District Energy Association）的 Rob Thornton 表示。“这能减少排放，降低人力成本……通过将功能集中在中央工厂，你可以为数十个甚至数百栋建筑提供服务。”

全球范围内的热浪正变得更强、更长且更严重，这推动了空调销量的激增。但在全球变暖最快的洲——欧洲，大约只有 1 in 5 的家庭安装了空调，而对于老龄化人口来说，缺乏凉爽空间可能是致命的。空调虽然对医院、学校和护理院至关重要，但它会加剧全球变暖的碳排放和城市热岛效应。根据地方政府联盟——欧盟市长公约（European Union Covenant of Mayors）2025 年的一份报告，空调并非长期解决方案。该组织表示，相比之下，区域制冷既能为城市降温，又能减少排放。

全年有数月面临酷暑的中东和亚洲城市已经开始建设巨大的区域制冷网络。迪拜的一家国家支持的区域制冷公司目前向 1700 多栋建筑输送约 6 gigawatts 的制冷量——相当于每天冻结 150 万 吨冰。

根据 Thornton 的说法，美国大约有 500 个此类区域制冷网络，主要分布在大学和医院校园内；欧洲大约有 200, 个，主要分布在法国和瑞典的城市中。这些制冷系统通常在制冷槽中将水冷却至 6℃ 或 7℃，然后将其通过管道输送到网络中每栋建筑的热交换器中。这些交换器的作用类似于反向运行的散热器，通过与冷水接触来冷却家中的空气。

香港的一项研究显示，一个由 37 栋公共和商业建筑组成的模拟区域制冷网络，与独立空调相比，能耗降低了 15%，运行成本降低了 10%。巴黎的区域制冷系统与独立制冷相比，电力消耗和排放降低了一半。

, Thornton 表示。区域制冷可以缩减那部分巨大的冗余容量。为了进一步提高效率，日内瓦和巴黎的区域制冷系统分别利用日内瓦湖和塞纳河的水来冷却其液体 “区域制冷系统的效率是两倍，或运行费用为一半”，而荷兰海尔伦（Heerlen）的一个系统则利用被淹没的煤矿。在美国，一些区域制冷网络在电价低廉的夜晚制造冰块，并在白天融化时将其输送出去。

虽然大多数区域制冷网络仍服务于公共或商业建筑，但少数网络（如日内瓦的网络）已开始扩展到住宅建筑。但区域制冷的成本可能

数亿美元的建设成本，这主要是因为铺设管道需要挖掘街道和人行道。宾夕法尼亚州立大学的左旺达（Wangda Zuo）表示，如果能与现有的区域供暖系统同步安装，成本将最低。他说：“这种方案适用于建筑密集且需求量大的地区，而最理想的情况是你已经拥有了铺设管道所需的隧道。”

二合一

但一种名为“环境环路系统”（ambient loop system）的新技术将区域供暖和供冷结合在一起，可能会在欧洲获得更广泛的采用。Zuo表示，一些包含商业和住宅建筑的高端项目已经安装了这项技术，但若要大规模普及，可能需要政府的支持。这种方法是一种增强个体电热泵性能的方式，而传统的电热泵通常是对房屋外部吸入的空气进行冷却或加热。环境环路系统循环流动的水温取决于管道当前的温度，通常在 10℃ 到 25℃（50°F 到 77°F）之间，尽管在冬季偶尔需要对其加热。然后，每栋建筑中的水基热泵会对这种液体进行冷却或加热。因此，例如在热浪期间，热泵无需将室外空气从 40℃ 冷却到 20℃（104°F 到 68°F），而是可以将水温从 25℃ 转换为 20℃，这所需的电量要少得多。“如果你想把东西从山脚搬到山顶，你必须非常努力地工作，”Zuo说。“如果高度差较小，就会更容易。使用环境环路，温差相对较小。” ■ 8 August 2026 / New Scientist / 13

---

为每个阶段提供支持

我们可以帮助您为今天以及接下来的任何挑战做好准备。 储蓄与抵押贷款

ybs.co.uk Trustpilot 10. 条评论 欢迎光临我们的分行

. 约克郡建筑协会（Yorkshire Building Society）已列入金融服务登记册，其注册号为 106085。 总部地址：Yorkshire House, Yorkshire Drive, Bradford BD5 8LJ. ybs.co.uk YBM 13893 17 07 26

---

新闻 健康

“药丸形式的低氧”可能缓解神经退行性疾病症状

Ute Eberle

一种模拟高海拔和低氧对大脑影响的药丸，似乎缓解了患有包括帕金森病在内的三种神经退行性疾病的小鼠的神经功能衰退。马萨诸塞州总医院的 Vamsi Mootha 表示：“通过‘药丸形式的低氧’，我们现在可以操纵我们呼吸的空气。”

虽然严重的高原反应可能是致命的，但轶事证据表明，像帕金森病这样的神经系统疾病症状在高海拔（空气较稀薄）地区会减轻。在过去的十年里，Mootha 及其同事进行了研究，将患有帕金森病等疾病模型的小鼠置于含氧量为 11% 的空气中，这与海拔约 5000 米的环境相匹配。他们发现，这可以停止、预防甚至逆转症状。

尽管这些小鼠在干预过程中没有出现任何已知的副作用，但长期全天候呼吸稀薄空气对人类来说是不安全的。Mootha 说：“鉴于低氧可能具有危险性，不应在“在帕金森病模型中测试时，小鼠的协调能力和平衡能力得到了改善”家中尝试。”

但现在，Mootha 及其同事将两种实验性药物结合，创造出了药丸形式的低氧。其中一种名为 osivelotor 的药物会改变血红蛋白，使其更紧地抓取氧气，从而阻止气体到达组织。但这会导致身体产生危险数量的红细胞（红细胞在体内运输氧气）。第二种药物 PT2399 通过阻断一种名为 HIF-2α 的蛋白质来防止这种情况，该蛋白质有助于细胞应对低氧。

研究团队将这种药丸给患有晚期 Leigh 综合征模型的小鼠服用，这是一种导致精神和运动能力逐渐丧失的严重神经系统疾病。该药丸将小鼠的生存时间从约 62 天增加到 158 天，增加了一倍多。

研究人员还在弗里德赖希共济失调（Friedreich's ataxia）模型小鼠中测试了该药丸，这是一种导致神经系统渐进性损伤并导致运动困难的罕见疾病。团队观察到，这停止了小鼠力量和协调能力的进一步下降。最后，在帕金森病模型中测试时，接受治疗的小鼠改善了协调能力和平衡能力 (bioRxiv, doi.org / rhrn)。

Mootha 表示，其效果与此类动物呼吸 11% 氧气时相当。研究表明，低氧可以增强大脑生长因子，从而有助于神经元之间形成新的连接。但 Mootha 提出，减少过量的氧气可能会防止细胞损伤。他及其团队关注的三种疾病部分特征是线粒体受损，而线粒体为细胞提供能量。他说，这些线粒体可能无法正确消耗氧气，导致局部氧气过剩，从而损害其他细胞成分。

佛罗里达大学的 Gordon Mitchell 表示，这项研究让他“非常感兴趣”，但提醒说，这些效果在人体上的转化情况仍有待观察。■ 动物行为

捷豹神秘跋涉 2100 公里

一只雄性捷豹完成了一次前所未有的跋涉，沿着亚马逊雨林的南缘行进了 2100 公里。这次旅程扩展了我们对这些大型猫科动物能力的认知。爱尔兰都柏林大学的 Jon Morant 表示，此前带标记捷豹的扩散记录约为 60 到 100 公里。作为一项始于 1999 年的持续项目的一部分，他及其同事一直在通过 GPS 项圈追踪巴西的捷豹。研究人员希望更好地了解捷豹如何利用阿拉瓜亚河沿岸的环境，该河流是连接塞拉多（Cerrado）草原与亚马逊的潜在野生动物走廊。作为这项工作的一部分，他们捕捉到了这只被命名为 Gaspar 的雄性捷豹，并在 2024 年 11 月 11 日为其佩戴了项圈。这只 8 岁的捷豹随后在其既定领地内停留了一段时间，然后开始了一次大规模的向南跋涉，沿着亚马逊与内陆塞拉多的边界行进。研究人员记录到 Gaspar 行进了 2122 公里， 佩戴着 GPS 项圈——直到 2025 年 11 月下旬项圈停止传输位置信息之前，它一直进行着这场令人惊讶的冒险（《生态学》 / Ecology, doi.org / rhqw）。Morant 表示，实际距离可能要远得多。虽然 主要留在森林地区，但它也穿过了被人类高度改造的地貌，例如农场、牧场和道路。对于 1 到 3 岁、试图建立新领地的个体来说，扩散（尽管距离较短）很常见，但目前尚不清楚像 这样年长的成年个体为何会采取这种行动。Morant 表示，它可能是在寻找更好的食物资源、交配机会或总体上更好的栖息地。■ Joshua Rapp Learn 2026 年 8 月 8 日 / New Scientist / 15 --- 新闻 考古学

扁平颅骨揭示铁器时代的育儿方式

Christa Lesté-Lasserre

扁头综合征通常与现代育儿习惯相关，尤其是让婴儿仰卧睡眠，而这种现象在至少 2800 年前可能就已经很普遍。由于人类出生时头部较软，颅骨容易受到压力的影响而变形。如今使用婴儿床和婴儿车往往会导致后脑勺承受长时间的压力，从而使原本圆润的表面变形。但扁头综合征（也称为体位性或发育性斜头畸形）的最早历史仍然未知。到目前为止，研究人员在巴基斯坦发现了两具可追溯至 4000 年前的斜头畸形颅骨，以及来自公元前 4 世纪瑞士和中世纪捷克共和国的案例。

罗马大学的 Michael Beck De Lotto 及其同事检查了 26 具保存完好的成年颅骨，这些颅骨来自 20 世纪 90 年代在意大利马特利卡（Matelica）发现的两处墓地，与公元前 8 和 9 世纪的皮切内人（Picenes）有关。“父母可能让婴儿仰卧睡眠，并用枕头垫高头部”

他们在 7 名男性和 4 名女性的共 11 个头部后方发现了扁平区域，这使其成为亚洲以外已知最早的发育性斜头畸形案例。研究人员对这些颅骨进行了 CT 扫描，并为每一个颅骨创建了 2D 和 3D 虚拟模型。De Lotto 表示，所有颅骨都没有显示出生前或死后的变形迹象，这意味着变形必须发生在头部形状固化之前的一岁之内。

与现代扁头综合征（主要集中在后脑勺）不同——CT 扫描突出了颅骨扁平的位置（红色部分）——这些铁器时代的颅骨具有多种扁平模式，代表了不同的体位（《国际古病理学杂志》International Journal of Paleopathology，doi.org / rhrq）。他说，头部形状表明，父母可能让婴儿仰卧睡眠，并用枕头垫高头部。他们也可能给婴儿包裹襁褓或将他们放在摇篮中，尽管具体机制尚不清楚。

德国儿科医生兼独立研究员 Herbert Renz-Polster 表示，这些发现与现有的关于古代育儿的“进化失配”理论相一致。他解释说，随着狩猎采集者转向流动性较低的生活方式，他们也从携带婴儿并共同睡眠，转变为一种将婴儿单独安置在床上并使用襁褓等“睡眠辅助工具”的育儿实践。这导致婴儿长时间仰卧不动，头部承受持续压力。■ 古生物学

恐龙被全球火暴毁灭

詹姆斯·伍德福德 (James Woodford)

根据撞击模拟显示，6600 万年前的小行星撞击引发了一场行星规模的火暴，将全世界的恐龙烧成了焦炭。所谓的希克苏鲁伯（Chicxulub）小行星在白垩纪末期落在如今的墨西哥湾地区。此前人们认为，热量和火灾最严重的影响相对局限于区域性，而全球性的灭绝主要是由“核冬天”引起的，即尘埃遮蔽太阳导致剧烈降温。

印第安纳州普渡大学的布兰登·约翰逊 (Brandon Johnson) 及其同事分析了地质观测和撞击模拟，结果显示，撞击导致汽化的物质比此前认为的多出约五倍。这项新研究表明，在撞击发生后，一股岩石蒸汽柱被喷射到数万公里外的太空，并最终分布在整个行星的高空。

小行星撞击引发全球范围的酷热冲击波

约翰逊表示，几秒钟后，这股蒸汽柱在地球大气层的大部分区域之上向太空扩张，导致尘埃以及被称为“球粒”的玻璃质金属小珠从蒸汽中凝结。研究发现，这团巨大的蒸汽云最终变成了1.6 trillion吨的尘埃，覆盖了整个行星，将热量封锁在内。较重的球粒则重新撞击大气层（《地球物理研究杂志：生物地球科学》Journal of Geophysical Research Biogeosciences，doi.org / rhq47）。

“随着这些物质在接下来的几个小时内重新进入大气层，摩擦力使球粒减速并升温，产生一个持续约一小时的热辐射脉冲，”约翰逊说。“这将杀死那些无法躲避的人，且强度足以点燃草和松针等森林地表覆盖物，甚至可能足以点燃树木。”他表示，地质记录表明，随后的火灾可能持续了数年。

“白垩纪动物接收到的热辐射剂量是人类100 per cent致死剂量的17倍，”约翰逊说。“这应该足以杀死即使是皮厚恐龙。” ■

16 / New Scientist / 2026年8月8日

---

我们想向您介绍我们的全系列产品

选择 Saga，您可以放心。我们在为50岁以上人群设计和提供屡获殊荣的产品方面拥有超过70年的经验，您可以信任我们提供高质量、高价值和卓越的服务。

邮轮

saga.co.uk / cruises

度假

.co.uk / holidays

保险

.. / insurance

杂志

.. / magazine

金融

.. / money

在 .. 了解我们全系列产品的更多信息 经验决定一切 NHA-AP6671

---

观点专栏 思考之食

越来越多的研究将超加工食品与抑郁、焦虑甚至痴呆症联系起来。海伦·汤姆森 (Helen Thomson) 深入探讨相关证据

海伦·汤姆森是包括《不可思议：一次穿越世界上最奇异大脑的非凡旅程》（Unthinkable: An extraordinary journey through the world's strangest brains）和《这本书可以改变你的生活》（This Book Could Fix Your Life）在内的多本书籍的作者。她此前曾任《新科学家》杂志的特写编辑。

海伦的这一周

我在读

泰勒·詹金斯·里德 (Taylor Jenkins Reid) 的《伊芙琳·雨果的七任丈夫》(The Seven Husbands of Evelyn Hugo) —— 完美的泳池边阅读之选

我在看

艾玛·巴内特 (Emma Barnett)：对抗子宫内膜异位症

我正在从事的工作

《Change Your Mind with New Scientist》第二季，这是我与《新科学家》编辑 Cat de Lange 共同主持的绝佳新播客。

直到现在，我们都知道超加工食品对我们的身体并不 particularly 有益。超加工食品（UPFs）通常含有高糖、高盐、饱和脂肪和添加剂，被认为与心脏病、糖尿病和某些癌症风险增加有关。但它们对我们的大脑有何影响？最近，我注意到越来越多的研究将超加工食品与抑郁、焦虑甚至痴呆症联系起来。我想看看这些证据有多强，以及它是否提供了另一个令人信服的理由，让我将它们从饮食中剔除。

目前直接的问题是，虽然许多观察性研究发现摄入更多超加工食品与各种大脑健康问题之间存在关联，但直接的因果证据较少。伦敦国王学院、营养公司 Zoe 的联合创始人 Tim Spector 表示，我们如何定义超加工食品使情况进一步复杂化。大多数国家使用 NOVA 分类系统，该系统旨在对加工水平进行分类，而非评估食品如何影响健康。

尽管如此，关于超加工食品危害的证据正在增加。2024 年，一项涉及约 3万,000 人且随访长达 18 年的研究表明，一个人饮食中超加工食品的比例每增加 1.0775万%，“多吃纤维和油性鱼类是你可以为大脑健康做出的两项最佳改变”，认知障碍和中风的风险就会显著上升。即使在考虑收入、酒精消费和体力活动等因素时，这种关联依然存在。

另一项研究对巴西的 10,775 名成年人进行了约八年的随访。那些从超加工食品中获取能量比例超过 19.9% 的人，其全球认知能力（一种衡量心理能力的综合指标）和执行功能（控制冲动、调节情绪和保持专注的能力）下降速度更快。

而且，这不仅仅是普遍的认知能力下降。2025 年，研究人员在分析 40 至 70 岁之间 5.8423万,423 人的数据时发现，超加工食品消费的增加与痴呆症、帕金森病和多发性硬化症的风险升高相关。其他研究也将超加工食品与抑郁和焦虑联系起来。在其中一项研究中，研究人员在 2003 年至 2017 年间对 3.1712万,712 名在研究开始时没有抑郁症的女护士进行了随访。那些每天食用超过 8.8 份超加工食品（如人工甜味饮料或预制餐）的人，患抑郁症的风险高出 49%。虽然是观察性研究，但该研究使用了一些先进的统计方法来排除“反向因果关系”——即如果你处于抑郁状态，可能会更倾向于食用不健康食品。

那么，超加工食品究竟做了什么，会对大脑产生如此糟糕的影响？答案可能涉及多个重叠的路径。例如，许多超加工食品含有高盐、高脂肪和高糖，这些都被已知会破坏微生物组，使其多样性降低且有益菌减少。这会损害肠道内壁，使其变得更“渗漏”，从而允许有害物质进入血液和大脑。

“高脂肪和高糖饮食还会诱发低度炎症，”Spector 说，“我们现在知道，炎症与心理健康，尤其是抑郁症之间存在密切联系。”挪威奥斯陆大学的 Laura Bojarskaite 向我提到了一项去年发表的小规模研究，该研究发现，仅在五天内通过超加工零食额外摄入 1500 卡路里，就会改变男性大脑对胰岛素（一种帮助细胞吸收葡萄糖以产生能量的激素）的反应。在他们恢复正常饮食一周后，这种影响仍然可以被检测到。

胰岛素调节对于维持认知功能至关重要，而胰岛素抵抗则与抑郁症和阿尔茨海默病相关。尽管该研究并未专门探讨大脑健康，但它提供了证据，证明超加工食品（UPFs）能够迅速改变大脑信号传导，博亚斯卡伊特（Bojarskaite）表示，这是一个“可信的路径，通过该路径，终身[高UPF]饮食可能会影响大脑衰老”。博亚斯卡伊特认为，与其试图从饮食中剔除每一种超加工食品，我们应该集中精力减少那些结合了多种不利特征（如添加糖、高盐和低纤维）的产品，并优先替换这些产品。如果这听起来仍然太难，爱尔兰科克大学（University College Cork）的特德·迪南（Ted Dinan）在我共同主持的播客节目《与《新科学家》一起改变你的想法》（Change Your Mind with New Scientist）最近的一集中提供了一个更简单的起点，他说，增加纤维和油性鱼类的摄入是你能为大脑健康做出的两项最佳改变。■ 18 / New Scientist / 8 August 2026

---

块状宇宙

宇宙学原理认为，宇宙中任何地方的所有事物在整体上都应该是大致相同的。利亚·克雷恩（Leah Crane）问道，现在是否到了放弃这一想法的时候？

在宇宙学中，有一个主导一切的原理：被恰如其分地命名为“宇宙学原理”。它坚持认为宇宙中没有特殊的地方——在足够大的尺度上，从所有视角看，所有方向上的所有事物都大致相同。这看起来一定是错误的：显然，宇宙中的某些地方与其他地方不同。例如，我的办公室与黑洞毫无相似之处。为了使其有用，我们必须缩小视角（zoom out）——这就是为什么现代的宇宙学原理表述中会加上“在足够大的尺度上”。任何一个超星系团都与任何其他超星系团大致相同。它是我们目前最优秀的宇宙模型（称为标准模型）的基石。即便如此，许多宇宙学家已经对这一基本法则提出了质疑。

该原理由两部分组成：均匀性（在足够大的尺度上，所有事物都是平滑的而非块状的）即使是龙虾星云，在足够远地缩小视角后看起来也是平滑的，以及各向同性（在每个方向上都是如此）。前者面临着更可信的威胁，这些威胁通常表现为尺寸异常的结构，导致其被命名为“巨型类星体群”或“巨弧”，所有这些结构的跨度都达数十亿光年。当宇宙中似乎存在如此巨大的块状物时，很难认为我们的宇宙是平滑的。插入在宇宙学原理中的“在足够大的尺度上”这一限定，其延伸范围也是有限的，对吧？

在 21 世纪初，天文学家开始发现一些规模稍微大到无法纳入标准模型的结构。并非所有人都同意界限应该在哪里，但提议的界限通常是几个百兆秒差距，或数亿光年——任何比这更大的结构都是一个问题，而确实似乎存在更大的结构。因此，整个宇宙的均匀性是值得怀疑的。

就个人而言，我并不太担心宇宙学原理本身可能失效。一些研究人员说，在标准模型中，少量的块状分布实际上是可以接受的；一些人说，宇宙自形成以来已经膨胀得如此之多，以至于我们实际上不能再指望宇宙学原理成立；还有一些人简单地表示，我们必须在更大的尺度上观察。但如果我必须缩小视角到如此程度才能使一个范式契合真实的宇宙，那么这个范式是否仍然有用？

宇宙学原理让我们能做出一个巨大且必要的假设：我们所处的位置并不特殊，我们在地球上观察到的现象在任何地方都成立。如果我们想继续研究整个宇宙，就必须假设某种程度的均匀性和各向同性。但仅仅因为一个假设有用，并不意味着它是真实的。

在科学领域，我们经常假设一些未必正确的事情，这在很大程度上是为了简化计算。我们测量到空间中任何物体的每一次距离（甚至包括宇宙学原理本身的距离尺度），都假设了一个宇宙膨胀率的数值，尽管我们并不完全清楚这个数值是多少。我们可以接受某件事在接近真相的程度上足以产生用处，同时深知它并非完全正确。那么，我相信宇宙学原理吗？不，并不真正相信。宇宙太庞大且多样性太丰富，以至于我无法完全认同“不存在独特之地”这个观点。至少，它足够无限，使得宇宙学原理可以既足够正确，又是不正确的。在我们宇宙的这个角落存在生命，因此其他地方可能也存在生命，因为在某些方面，宇宙在任何地方都是相同的。但它在任何地方也都是不同的。存在特殊之地，而我们正处于其中之一；只有将这一点与宇宙学原理同时铭记在心，我们才能真正理解我们在宇宙中的位置。■

莉亚·克雷恩（Leah Crane）是《新科学家》杂志的高级记者，报道的主题涵盖从私人太空飞行、宇宙学到量子力学等领域

Leah 的一周

我正在阅读 塞缪尔·泰勒·柯勒律治的《古舟子咏》 我正在观看 出于某种原因，在看大量的豹子重播节目 我正在忙于 准备前往北极观看日全食！

LIGHT, BIRARD, MALAYSIA, CALTECH 2026年8月8日 / New Scientist / 19

观点专栏作家 针对“不可成药”蛋白质的药物 Michael Le Page 表示，一种运作方式完全不同的新型药物已获得批准，它可能会彻底改变治疗方案。

Michael Le Page 是《新科学家》的记者，专注于遗传学、生物医学和环境领域。

Michael 的一周 我正在阅读 我刚读完泰勒·詹金斯·里德的《Atmosphere》 我正在观看 我终于开始在 Netflix 上看《恐龙》了 我正在忙于 我已经开始为下一篇专栏（可能是关于肌肉的）采访相关人员了

我认为，当我们回顾 2026 年时，会将其视为医学的一个里程碑，因为这一年第一款 PROTAC 药物获得了批准。大多数药物只是简单地阻断其目标，而 PROTAC 则能将目标彻底摧毁。就单个分子而言，PROTAC 的效力要强得多，而且它们能完成其他药物无法实现的操作。

要理解为什么 PROTAC 如此具有革命性，我们需要从传统药物说起。你服用的几乎所有药片都包含由小分子组成的药物。这些分子的形状使它们能够与特定蛋白质结合，从而阻断或刺激它们。小分子药物具有一些显著优势：制造成本低，能够进入细胞，并且可以通过肠壁进入血液，因此可以作为药片服用。

但它们体积小，意味着特异性不高。它们倾向于与目标分子以外的其他分子结合，从而产生不理想的副作用。而且在许多情况下，根本无法找到能阻断蛋白质活性的小分子。问题在于，每个药物分子不能随便结合目标蛋白质的任何部分。为了产生效果，它必须结合在蛋白质的活性位点——即“工作面”上——但有时活性位点的小部分区域没有足够独特的形状。

“在临床试验中，接受 PROTAC 药物治疗的晚期乳腺癌患者生存时间延长了两倍以上”

这类蛋白质被称为“不可成药”的。这就是为什么许多较新的药物由大分子组成，最常见的是被称为抗体的蛋白质（任何名称以“-ab”结尾的药物都是抗体）。抗体较大的体积意味着它们能以极高的精度结合目标，使其比小分子药物更有效且更安全。要制造抗体，你需要让免疫细胞针对所需目标产生抗体，然后挑选出最佳的一种。这听起来很简单，但花了数十年才得以完善。

抗体的缺点是价格极其昂贵，因为它们只能在活细胞中制造。其体积较大也意味着必须通过注射给药，因为它们无法通过肠道。而且一旦进入血液，它们也无法进入细胞。因此，虽然抗体可以结合许多不可成药的蛋白质，但它们无法进入细胞去触达这些目标。

另一件需要提及的是，小分子药物和抗体都必须与目标接触才能发挥作用。PROTAC 的运作方式则不同。本质上，它们在细胞内的蛋白质上贴了一个“回收我”的标签。这意味着仅一个 PROTAC 分子就能标记并清除数十个或数百个该蛋白质分子。而且因为它们不是在阻断或刺激目标蛋白质，所以不需要结合在活性位点。目标蛋白质上的任何位置都可以，这使得许多不可成药的目标变得可成药。

美国食品药品监督管理局（FDA）在 5 月批准了一款用于治疗乳腺癌的 PROTAC 药物。

制造足够小且有用的protac分子是一项挑战，因为每个protac必须结合两个分子，而传统的小分子药物仅需结合一个。Craig Crews与Raymond Deshaies在1998年共同提出了protac的想法，他成功地将它们缩小到能够像小分子一样运作的程度，但它们仍然比传统药物略大。“我称它们为‘大型小分子’，”Crews上个月告诉我。到2013年，Crews有了足够的信心创立了一家名为Arvinas的公司。在开发首批产品时，Arvinas采取了稳健策略。它没有直接挑战那些不可成药的蛋白质，而是将目标定为癌细胞中的激素受体。例如，某些类型的乳腺癌生长是由对雌激素有反应的受体驱动的。目前已经有几种通过靶向这些受体起作用的药物，但可能会产生耐药性。癌症产生耐药性的一种方式是制造更多的受体，直到即使是最高可耐受的药物剂量也无法阻断所有的受体分子。“由于我们是在主动降解受体，所以我们没有这个限制，”Crews在2020年告诉我。在临床试验中，与使用传统药物的患者相比，患有已扩散至全身的晚期乳腺癌的患者在接受protac药物vepdegestrant治疗后，生存时间延长了两倍多。因此，vepdegestrant在5月获得了美国食品药品监督管理局的批准。“这次批准是该领域的一个里程碑，”Crews上个月告诉我。我希望很快能看到各种基于protacs的新疗法，最终造福我们数十亿人。vepdegestrant的获批可能几乎没有引起关注，但如果它确实是众多药物中的第一款，我们回头看时会将其视为一个重大里程碑。 ■ 20 / New Scientist / 8 August 2026 --- New Scientist 订阅

半价优惠 一年 £99

在这个夏天释放你的好奇心，通过无限量的数字访问权限，随时随地探索最新的科学观点。立即在 newscientist.com / summersale 订阅。如需通过电话订购，请致电 +44 (0) 330 333 9470，并提供代码 22393。这些是自动续订的订阅，可随时取消。我们在首次付款后提供 14 天的冷静期。优惠于 2026 年 10 月 06 日结束。

---

观点 视角

22 / New Scientist / 2026 年 8 月 8 日

---

净损失

Gaurav Patil 红树林摄影奖

这张不同寻常的照片是在紫外线下以 13 秒曝光拍摄的，展示了海上捕鱼对陆地造成的部分破坏。Gaurav Patil 拍摄了一张渔网的照片，该渔网被洋流携带，缠绕在印度孟买以南南德高村岸边的一棵红树上，这张照片赢得了 2026 年红树林摄影奖的总冠军。该比赛由美国非营利组织红树林行动项目组织，旨在提高人们对全球红树林处于危险境地的认识。红树林占所有热带森林的比例不足 1%，但为野生动物提供了至关重要的栖息地，且其碳存储量最高可达同等面积内陆森林的 10 倍。尽管在经历了数十年的衰退后，近期出现了一些好消息，但据认为超过一半的红树林生态系统在 2050 年前面临崩溃风险。成千上万吨丢失、遗弃和丢弃的渔具散落在海洋中，被洋流带到全球各地，最终被缠住或分解成微塑料。这些渔具会严重破坏红树林生态系统，缠绕海洋生物并使红树林窒息。“这位摄影师赋予了一个通常看起来枯燥但对保护至关重要的话题以生命。幽灵网是一个真实存在的问题，”评委 Dhritiman Mukherjee 在颁奖公告中表示。■

Matthew Sparkes 2026 年 8 月 8 日 / New Scientist / 23

观点 本周之书

生命的难题

一本极具挑衅性的书籍试图以截然不同的方式来描述生物世界——它坚持认为， agency（能动性）和目的性无处不在。Michael Marshall 评述 Melanie Challenger 的著作 《活着：生物世界的隐藏智能》 (Alive: The hidden intelligence of the living world)，由 Canongate (英国) 和 Penguin Random House (美国) 出版。

当你思考这个问题时，活着其实是一种非常奇怪的存在状态。宇宙中绝大多数的物质是毫无生命的氢和氦、岩石和尘埃。然而，不知为何，我们现在都在这里：数以百万计的微生物、植物和动物物种，都在呼吸、进食并试图繁衍，其中一些在尖叫，一些在睡眠，而我们最终都会死去。生物化学家雅克·莫诺（Jacques Monod）在《偶然与必然》(Chance and Necessity) 中写道，当我们考虑到生命在宇宙中出现的极低可能性时，“我们应该感到奇怪，且觉得有些不真实”。

实际上，定义什么是生命已被证明极其困难，因此我们倾向于求助于隐喻。过去一个世纪以来最著名的隐喻之一就是：生物有机体是机器。它们拥有组件、活动部件以及为了促成某些结果而存在的复杂化学信号通路。在这种观点中，即使是大脑也是一种计算机，一种对刺激做出反应并进行计算的计算机。此外，这些机器拥有由进化塑造的目的。在《自私的基因》(The Selfish Gene) 中，理查德·道金斯（Richard Dawkins）认为，有机体是旨在传递基因的机器，生物仅仅是基因自我复制过程中产生的复杂副产品。

在《活着：生物世界的隐藏智能》(Alive: The hidden intelligence of the living world) 一书中，作家兼广播员梅兰妮·查伦杰（Melanie Challenger）旨在彻底摧毁这种思维方式。她认为，生物不是机器，它们的存在也不是纯粹为了传递基因。相反，定义生物的是能动性（agency）和目的性。她写道：“生命形式追求目标是因为它们活着，而生命在本质上就是有目的的。”

查伦杰认为，认为有机体拥有目标和目的，是对“生命是什么”这个问题的“最显而易见”的答案，同时也是“最被奇怪地忽视”的答案。从勒内·笛卡尔（René Descartes）开始，非人类动物拥有自己的目标、感受甚至思想这一观点，一直受到动物学家和心理学家的质疑。动物行为学家 C·劳埃德·摩根（C. Lloyd Morgan）曾提出著名的论点：如果某种动物行为可以用更简单的过程来解释，那么我们绝不应将其解读为更高阶心理过程的证据。尽管“更高阶”和“更低阶”的心理过程具体指什么并不明确，但“摩根法则”（Morgan’s Canon）随后指导了一个世纪的研究。

查伦杰的建议是，我们无法摆脱“自我”这一概念（即某种具有能动性的东西）来解释生物有机体的行为。只不过我们将它从有机体本身中移除了：要么将其表现为某种可分离的灵魂或本质，要么将其表现为一套决定有机体如何为了自身目的而行动的基因。“灵魂和基因都被描述为类自我的代理者，是生命形式行为的真正发起者，”她写道。在这些世界观中，“生命形式本身被边缘化了”。

相反，她认为我们应该将生物视为“自我决定”的：“生命形式利用能量来保持与环境的局部不连续性……而这一行为的非凡之处在于它是自我生成的。”

查伦杰在六个章节中展开这一观点，探讨了稳态（有机体维持内部环境的机制）、感觉感知、意识和智能。这些现象虽然都可以利用细胞、分子和物理力进行机械论的解释，但只有在整体有机体的视角下，它们才具有意义。

查伦杰试图夺回生命的能动本质。即使是像植物这样没有大脑、因此可能几乎没有或完全没有意识体验的生物，仍然能对环境做出反应，并为了自身的目的而有目的地行动。动物或植物没有任何像自动机（automaton）一样的地方：正因如此，科幻小说中的自动机在对比之下才如此深刻地、在直觉层面上令人不安。开诚布公地说：我认为

关于生命意义的三本推荐读物

《生命的边缘：寻找生命的定义》（Life’s Edge: The search for what it means to be alive），作者：卡尔·齐默（Carl Zimmer）。尽管经历了数百次尝试，我们仍然没有一个统一的生命定义。科学作家卡尔·齐默探讨了许多棘手的边缘案例，例如病毒，它们模糊了生命与非生命之间的界限。

《共生：共生关系与生命最伟大协作的隐藏故事》（Togetherness: Symbiosis and the hidden story of life’s greatest collaborations），作者：罗文·胡珀（Rowan Hooper）。作为《活着》的理想配套读物，罗文·胡珀（《新科学家》播客编辑）向我们展示了将生命理解为一个共同体的重要性。个体生物仅仅是故事的一部分。

《自由代理人：进化如何赋予我们自由意志》（Free Agents: How evolution gave us free will），作者：凯文·米切尔（Kevin Mitchell）。神经科学家凯文·米切尔认为，生物是“有目的行动的代理人”，而不仅仅是对刺激做出反应的机器。他的观点正是梅兰妮·查伦杰的灵感来源之一。 24 / New Scientist / 2026年8月8日 ---

查伦杰在很大程度上是正确的。从基因视角看待进化会误导我们，因为它赋予了基因过多的能动性和权力。实际上，基因并不是细胞的主控制器，也不包含整个生物的蓝图：它们是支撑生命的相互作用系统网络的一部分。基因在细胞中很重要，但其控制力并不比其他任何组件更高。我可以想象《活着》会有两类读者，且反应截然不同。一种合理的反应是，这简直显而易见。动物出于纯粹的生命力、饥饿、快乐和生存意志而有目的地行动，这一点是非常清晰的。问题在于，“生命即自动机”的观点已经获得了巨大的影响力。查伦杰认为，这一观点伴随着工业化农业的兴起而发展，而工业化农业在惊人的规模上剥削和虐待动物：如果你不完全相信动物具有感知力，就更容易接受这种全球性的虐待。如今，合成生物学家经常将生命视为字面意义上的可编程物质：他们是第二类读者，可能会发现查伦杰的立场难以接受。我认为查伦杰确实面临一个主要困难：她的论点集中在个体生物上。无论是细菌还是鲨鱼，每个个体都为了自己的目的而行动。她

一个巴西热带雨林支撑着数万亿个个体生物。查伦杰正确地强调了个体具有巨大的能动性，但她发现自己不得不忽视或淡化这样一个事实：个体的界限是模糊的，且个体只能在生态系统中生存。以你的身体为例。是的，你是你，你为自己而行动。但你也是数万亿个微生物的家园，它们影响着你，而其中一些微生物是你生存不可或缺的。查伦杰坚持将微生物视为与你截然不同的个体；每一个微生物都遵循自己的目标，而它们的目标恰好在大多数情况下与你的目标一致。但我认为，这过于简化了。在对待真社会性昆虫时，她也遇到了类似的问题——在这种昆虫中，只有一个个体进行繁殖，巢穴中的其余个体则是绝育的。真社会性是那些至少在某种程度上需要在基因层面进行解释的现象之一。查伦杰不喜欢将真社会性蚂蚁群体标记为“超个体”，但她的理由很牵强。她认为工蚁仍然有能力反抗系统，因此它们并非完全融合，但同样的情况也发生在我们的某些细胞身上——这就是为什么我们会得癌症。我的另一个提醒是，她那些高深的比喻和哲学术语并不适合所有人。这里有一个例子：“生物的活动是我们对世界特征之理解的字面体现，由生物自身从根本上决定。”如果你对此类表述的耐受度较低，可能会觉得这本书很难读，但这份努力是值得的：《活着》将帮助你以全新的视角看待生物世界。■ Michael Marshall 是一位常驻英国德文郡的科学作家

《新科学家》推荐

Eleanor Parsons

杂志编辑 伦敦

凝视着这具化石骨架，我看到了非同寻常的东西。在这一具 1.82 亿年前鱼龙的肋骨之间，有另一个更小爬行动物的遗骸，可能也是一只鱼龙——一场同类相食的晚餐，或者是一次化石化的妊娠。该标本作为“侏罗纪海洋：深海怪物”展览的一部分，在伦敦自然历史博物馆展出至 1 月 3 日。此次展览利用化石、铸件和视频，讲述了一些有史以来最凶猛的海洋掠食者的故事。这里有来自一种类似鳄鱼的短颈ب利奥龙（pliosaur）的大颚。你甚至可以触摸到沧龙（mosasaur）的牙齿（其头骨如图所示），它生活在稍晚的白垩纪。如果你好奇它的触感，它是尖的但很光滑，带有表示生长线的缓坡。我非常喜欢那只鱼龙的眼睛，以及来自鱼龙或蛇颈龙的、布满鱼骨的粪便化石。但我最喜欢的是那只化石化的天使鲨，看起来它仿佛仍躺在海底，等待着猎物。

伦敦自然历史博物馆信托基金会 2026 年 8 月 8 日 / New Scientist / 25

---

观点 文化

走进未知

一部关于太空探索成功与失败的深度历史回顾，提醒着我们所有离开地球的人将面临什么，Karmela Padavic-Callaghan 著

[...OMITTED...]

Out of This World

Peter Bond；Reaktion (英国)；芝加哥大学出版社 (美国)

俄罗斯宇航员和美国宇航员穿太空服的方式是否不同？在国际空间站 (ISS) 上演奏的第一件乐器是什么？谁是第一个将腌黄瓜走私到轨道上的人？

1957年，苏联发射了第一颗人造卫星 Sputnik 1。从那时起，人类通过直接前往太空来探索宇宙，这一壮举充满了挑战、悲剧和古怪之事。在《Out of This World: Life as an astronaut》一书中，作家 Peter Bond 详细记录了其中的许多故事，对太空旅行和太空生活提供了一个见闻广博且不事雕琢的视角。

他的书是以主题而非时间顺序编排的，第一章恰如其分地命名为“学习离开地球”。这一章概述了早期的太空飞行，以及使之成为可能的火箭和飞机的开发，例如第二次世界大战时的 V-2 火箭——它是第一代太空发射器的基础。Bond 指出，太空探索一直与大国之间的权力角逐紧密相连。这一部分本可以通过增加更多先驱者的轶事和个人细节来丰富。相反，《Out of This World》节奏很快，涵盖了极广的内容。

Bond 接下来的关注点是各国航天机构（如 NASA）筛选宇航员的过程，以及他们必须接受多么艰苦的训练。俄罗斯宇航员在星城（Star City）生活和训练，那是位于莫斯科东北部的一个独立小镇，而 NASA 的招募人员主要在德克萨斯州休斯顿接受训练。

如果有人曾在太空飞行，你大概能在《Out of This World》的某个地方找到他们的名字。我们还能听到宇航员本人的心声，因为 Bond 引用了他们的日记和采访，为他的书注入了必要的人文火花。

。

“自从亚当以来，还没有人像迈克·柯林斯在绕到月球背面时那样感受到如此深沉的孤独，”他在任务日志中写道。

宇航员在工作中也会像其他人一样变得暴躁。例如，Bond 透露在 1968 年，由于不小心将感冒带上了飞船导致鼻子堵塞，Apollo 7 的机组人员就因为是否要戴头盔而与任务控制中心发生了争吵。

虽然 Bond 主要讨论美国和俄罗斯的太空任务，但中国的航天员也经常成为关注焦点，这或许预示着一个更加国际化、但政治局势绝不会因此变得轻松的未来。这些紧张关系在 Bond 的倒数第二章中尤为突出，该章探讨了登月任务。冷战时期将人类送上月球的竞赛是美国胜过苏联的一个例子，但太空探索从未在政治上变得简单。美国试图让人们重返月球的努力就证明了这一点。

从左起顺时针方向：2026年 Artemis II 的发射；1970年的 Apollo 13 机组人员；首位进入太空的女性瓦伦蒂娜·特列什科娃；Bruce McCandless II 在 1984 年进行的无绳太空行走

在撰写本书时，Bond 承认 NASA 旨在完成该任务的 Artemis 计划仍有许多悬而未决的问题。在我阅读《Out of This World》时，该任务的一部分已凯旋完成，而另一部分实际上经历了重组。值得注意的是，Bond 几乎没有讨论关于为了私人和国家利益而殖民月球和火星的争论。

本书以庄严肃穆地回顾所有在太空丧生的人员作为结尾，从广为人知的事件——如 1986 年导致 7 名机组人员死亡的挑战者号灾难——到

26 / New Scientist / 8 August 2026 ---

回溯古代历史

汤姆·希德勒斯顿（Tom Hiddleston）成了庞贝纪录片出人意料的主持人。乔什·贝尔（Josh Bell）认为，这部片子虽然毫无掩饰地矫情，但不知为何却奏效了。

早期俄罗斯宇航员中的遇难者。邦德（Bond）的文字从未具有抒情色彩，但在最后一章中，少许情感悄然渗入：他写作是为了纪念那些逝去的生命，并解释他们的上级和支持团队是如何辜负他们的。如果 350 页耐心地剖析太空探索极具挑战性的诸多方式仍不足以遏制人类残存的傲慢，那么提醒我们面对宇宙之浩瀚时是多么脆弱和易逝，或许能起到作用。下一次前往国际空间站（ISS）的载人飞行计划于下月进行。想到这一点，以及《Out of This World》中所记录的漫长航天史，我会感到敬畏且焦虑。感谢邦德，我还会记得那些令人愉悦的故事，比如俄罗斯宇航员伊戈尔·沃尔克（Igor Volk）在 1984 年 7 月的联盟号 T-12 任务中偷偷带上了腌黄瓜。■

电视 《庞贝：时空之外》（Pompeii: Out of Time） Disney+

2023 年，当询问男性多久思考一次罗马帝国的病毒式趋势席卷社交媒体时，显然是一个错误——没有人向演员汤姆·希德勒斯顿提出这个问题。他显然经常思考罗马帝国，以至于现在制作并主持了一整套关于它的三集国家地理纪录片。

在《庞贝：时空之外》的第一集开头，希德勒斯顿讲述了 1998 年青少年时期访问这座保存完好的古罗马城市时所受到的影响，并展示了一些年轻时的迷人照片；在那种旅行中，其他人可能会把时间花在喝酒和派对上。相反，希德勒斯顿说，他在庞贝的经历对他产生了如此深远的影响，以至于引导他在大学学习古典学和拉丁语，甚至决定成为一名演员。

虽然他在《时空之外》中咨询了多位学者，但该节目发挥了希德勒斯顿作为表演者的能力，以提供对那些生活在公元 79 年被维苏威火山爆发摧毁的城镇中的人们的洞察。“我需要将历史证据与我作为演员所做的工作结合起来，”他说道，语气中带着一个对待其工艺极其认真的人所特有的庄重。

然而，希德勒斯顿那种毫无掩饰的矫情正是让这部剧奏效的原因。在介绍他的第一位专家之前，他吹捧一项革命性的新技术，可以追踪庞贝的实际居民，结果发现那只是古典学教授史蒂文·L·塔克（Steven L. Tuck）创建的一个简单的电子表格。随后，希德勒斯顿召唤出特效，将塔克的数据转化为三维全息图，并俏皮地注意到：“我看过一些《复仇者联盟》电影。”

他在漫威电影宇宙中饰演洛基（Loki）的经验继续影响着《时空之外》，前漫威影业高管凯文·R·赖特（Kevin R. Wright）也是该片的制片人之一。希德勒斯顿用大拇指和食指的一个小动作来模仿洛基，这个动作在整个系列中多次出现，用于在加入新信息时暂停或回溯瞬间。

该系列将旁白、采访与戏剧性的重演相结合，希德勒斯顿经常直接出现在扮演古罗马人的演员身边，就像洛基穿越时空见证历史事件一样。这些戏剧化演绎的演技比典型的历史重现更好，而且比导演保罗·W·S·安德森（Paul W.S. Anderson）那部夸张的 2014 年动作电影《庞贝》更接地气，而后者是最近一部高知名度的……

地质学家克里斯·杰克逊（Chris Jackson，左）与汤姆·希德勒斯顿在维苏威火山相遇

这一主题的流行文化解读。导演汤姆·巴博-迈特（Tom Barbor-Might）和剧集编剧创造了希德尔斯顿所追求的情感共鸣，他挑选了三个特定的真实人物作为焦点：青少年铁匠学徒阿维亚努斯（Avianus，弗兰基·特雷德威饰演）、富有的女商人朱莉娅·费利克斯（Julia Felix，鲁布娜·阿扎巴尔饰演）以及一名退役士兵，即禁卫军（Praetorian，巴里·艾尔德饰演）。起初，希德尔斯顿暗示这种方法将揭示关于实际发生之事的真相，但当然，这并不比诡计之神洛基说过的话更真实。尽管如此，这并没有让《Out of Time》变得不那么有趣或缺乏信息量，尤其是对于那些不像社交媒体上的那些男人那样频繁思考罗马帝国的人来说。通过利用历史学家、考古学家、地球科学家和一名“灾难心理学家”的专业知识，《Out of Time》阐述了庞贝事件的科学和文化背景。希德尔斯顿做了很多若有所思地凝视远方的动作，而且他与专家之间显然经过剧本设计的闲聊偶尔听起来有些生硬。当他第一次展示那个贴满文件的软木板时，很难不笑出来，就像他在进行一场谋杀案调查一样。当他的“办公室”随着庞贝地震的轰鸣而摇晃并崩塌时，情况变得更加滑稽。但希德尔斯顿也对该主题表现出了真诚且具有感染力的热情，而这一点起到了很大作用。如果他必须通过提及几次漫威（Marvel）来吸引人们的注意力，那么这是很小的代价。■ 乔什·贝尔（Josh Bell）是一位常驻美国拉斯维加斯的作家和电影评论家。

8 August 2026 / New Scientist / 27 --- 第二轮发售门票最高可省 47% 第二轮发售于 13 AUGUST 结束

NewScientist Live

立即订票，享受优惠 4 个阶段 50+ 位演讲者 展会各处设有互动体验 10 – 11 十月 学校日 12 十月 伦敦 EXCEL 中心 + 在线 以最优可用价格锁定您的门票 newscientist.com / nsl2026 荣幸获得 Yakult、THE HUSON CLUSTER FARMERS、伦敦国王学院（KING'S College LONDON）、医学研究基金会（Medical Research Foundation）、技术人员（Technicians）We make the difference、THE PAPER 的支持 47% 的折扣适用于周末家庭票，亦提供其他票种选项。如果您是《新科学家》的付费订阅者，请在结账时输入您的 12 位订阅编号（包括任何零），即可享受当前票价 10% 的折扣以及在线升级 50% 的折扣。

---

观点 您的来信

编辑精选

限制开源软件无法阻止滥用

社论，7 月 25 日 来自 Sam Edge，英国汉普郡，林伍德

您似乎在批评开源软件，理由是它能让每个人接触到潜在的危险技术。然而，密码学中有一句话同样适用于此处：“通过隐匿实现的安全性根本不是安全性。” 就像试图限制人们接触化学和弹道学书籍无法阻止不法分子大规模杀人一样，试图禁止开源控制系统或大语言模型也同样无效。无论如何，闭源软件也可以被逆向工程，其源代码可以通过工业间谍活动被窃取，或者简单地被盗版并用于非品牌用途。

如何适应高温，来自一名佛罗里达人的建议

7 月 18 日，第 16 页 来自 Craig MacDougal，美国佛罗里达州，坦帕

Alec Luhn 在热室中经历的条件，在我居住的佛罗里达州的夏季下午被认为是正常的。这里有一个来自亚热带气候的建议：要感知某一天会多么难受，请关注露点，而不是相对湿度。后者的数值在早晨较高，随着温度升高而下降，可能会给人一种下午感觉会好转的印象。（其实不会。）而露点——即相对湿度为 100% 时的温度——则会保持相对稳定。一旦露点升至约 21℃ (70°F)，无论气温如何，都会感到潮湿。我计算出 Alec 测试时的露点约为 23℃ (73°F)。露点达到 25℃ (77°F) 或更高（这在当地经常发生）会让人极其难受，甚至可能产生危险。我于 1978 年从佛罗里达州东北部的杰克逊维尔搬到坦帕。我已经适应了。在 69 岁这个年纪，我每周打两次网球，喝运动饮料，并大量出汗。并非每个人都能做到这一点，但我只要能行就会一直坚持。

夏克尔顿的照片暗示了一场被遗忘的不公正

7月18日，第18页 来自英国爱丁堡的玛格丽特·库斯伯特（Margaret Cuthbert）

你们最近刊登了一张极具戏剧性的照片，照片中是“詹姆斯·凯尔德号”（James Caird），这艘船在1916.年载着欧内斯特·夏克尔顿（Ernest Shackleton）及其五名船员穿越南极海域前往南乔治亚岛。这张照片同时也凸显了一场长期存在的不公正。在失去他的船只“坚忍号”（Endurance）后，夏克尔顿计划让船员们在冰雪上拖曳三艘救生艇，直到抵达开阔水域。这一计划遭到了木匠哈里·麦克尼什（Harry McNish）的沮丧反对，因为他预见到在崎岖地形上拖曳很可能会对船只造成严重损坏。据报道，这场争吵以夏克尔顿向麦克尼什举枪并赢得争论而告终，但在两天之内，他就不得不放弃计划并返回基地。最终，当冰层破碎时，他们驾驶三艘救生艇驶向象岛。正是在那里，麦克尼什对三艘船中最大的一艘——“詹姆斯·凯尔德号”进行了改装，使其具备了前往南乔治亚岛的航海能力。你们刊登的照片在一定程度上表明了麦克尼什的改装规模之大。这项工作是在极寒环境下、用冻伤的手指完成的，是对探险者们生存至关重要的贡献，而且只有麦克尼什拥有完成此项工作的技能。然而，在后来颁发极地奖章时，由于夏克尔顿未能推荐他，麦克尼什被剥夺了应得的荣誉。这是一场现在纠正还不算太晚的不公正。

智能生命不会像我们这样渴望权力

7月4日，第41页 来自英国坎布里亚郡安赛德的戴安·西尔维斯特（Dyane Sylvester）

布莱恩·拉基（Brian Lacki）在寻找地外智能时表现出了一种典型的拟人化偏见，他假设任何“先进文明”都会开发计算机并需要巨大的电能。一个真正智能的物种，难道不会使用仅需极少电能和资源的简单技术，以确保其能够在母星的承载能力范围内生存吗？

经济与环境难以兼顾

读者来信，7月18日 来自英国萨里郡卡特汉姆的蒂姆·穆尔斯利（Tim Moulsley）

正如基思·诺布尔（Keith Noble）在信中所述，澳大利亚政府为保护大堡礁而投入的资金无疑值得赞赏。然而，如果结论是基于基思提到的——该珊瑚礁支撑了7.7万,000个就业岗位并每年为国家经济贡献AUS90 亿美元 billion——那么在得出这是一个非常好的结论之前，人们应当谨慎。除非该经济活动严格实现净零排放，否则它将直接导致气候变化，而气候变化正是最初促使采取保护措施的主要因素！我还想知道，这些活动中有多少实际上可以被归类为乔尔·米尔沃德-霍普金斯（Joel Millward-Hopkins）在《过上美好的生活》（7月18日，第32页）一文中提到的“无用的过度消费”，因此在我们努力确保每个人拥有合理的基本生活质量且保持在地球安全运行空间内时，应该是需要被遏制的部分。我充分意识到珊瑚礁对旅游业的巨大吸引力，但很难看出这会有长期的未来，除非作为所有其他人类活动总和的一部分，拟议中的“大堡礁经济区”能变得真正可持续。

谈到大脑健康时，不要忘记平衡能力

7月11日，第32页 来自澳大利亚墨尔本的黄薇（Wai Wong）

一个经常被忽视的大脑健康衡量指标是平衡能力。虽然一般的锻炼对大脑健康有益，但针对平衡、力量和动作的训练方案（如太极）比仅关注单一领域的方法更能改善认知功能。互联网上可靠的网站提供了各种形式的简单且免费的平衡测试。因此，在你的大脑检查和锻炼计划中加入平衡训练是值得的。

英国地热能的另一种选择

7月25日，第32页 来自 Larry Stoter，英国蒙茅斯郡，北部 你们关于新地热方法的文章没有提到利用废弃矿井。在英国，有许多此类前煤矿可以被用作热源。虽然温度可能不足以发电，但对于区域供暖方案来说已经足够高。此类方案在盖茨黑德已经投入运行。

特此说明

冰岛的蓝湖是一个人工地热水疗中心（7月25日，第32页）。

想要联系我们？

请将信件发送至 letters@newscientist.com；条款详见 newscientist.com / letters。发送至 New Scientist, 9 Derry Street, London, W8 5HY 的信件将会延迟。

2026年8月8日 / New Scientist / 29

特写

封面故事

ANDREA DE SANTIS 30 / New Scientist / 2026年8月8日

---

预测机器

一种新型的人工智能正开始理解物理世界中的因果关系，丹尼尔·科辛斯（Daniel Cossins） 发现。

当你把一个玻璃杯从桌上撞下去时会发生什么？询问 ChatGPT，你会得到一个逐步的解释：重力如何让容器沿着“抛物线轨迹”加速，以及为什么如果“冲击产生的压力超过了材料所能承受的限度”，它可能会破碎。询问我 3 岁的女儿，她不太在意这些细节：“它碎了，水到处都是！”

在谈到人工智能的未来时，这种差异具有启发意义。驱动当今聊天机器人的大语言模型（LLMs）是在海量文本上训练的，旨在预测序列中的下一个词。然而，与小阿德琳不同，没有任何一个 LLM 真正地把玻璃杯从桌上撞下去过——或者向墙上扔过意大利面，或者把踏板车开进池塘里——而这可能会给它们的能力带来根本性的限制。

这基本上就是为什么一些研究人员认为，要将 AI 提升到下一个水平，我们需要的不是规模越来越大的 LLMs，而是“世界模型”：通过观察学习，从而能够模拟现实世界中行动后果的系统。

世界模型已迅速成为 AI 的下一个前沿，由于在自主机器人领域的潜力，吸引了工业界和商业界的巨大关注。然而，世界模型方法之所以特别引人入胜，是因为它表面上也提供了一条通往通用人工智能（AGI）的路径，即拥有人类水平推理能力且可应用于各种任务的机器。

问题在于，很难将实质内容与吹嘘区分开来。首先，“世界模型”这个词的含义模糊且具有弹性——对某些人来说，这种弹性甚至成了问题。新墨西哥州圣菲研究所的梅兰妮·米切尔（Melanie Mitchell）表示，它已成为“一种对当前 AI 系统无法良好完成的所有事情的简写”。目前尚不清楚这些系统究竟应该对什么进行建模，更不用说物理现实的内部表示是否足以实现那种泛化智能——即我的 3 岁女儿即使不能解释原因，也能直觉地理解她扫掉一个玻璃杯时会发生什么。

要理解 AI 中应用的世界模型，有助于了解该概念在认知科学中的根源。AI 研究人员通常将其追溯到 1943 年，当时心理学家肯尼斯·克雷克（Kenneth Craik）写道，人类心智“携带一个外部现实及其自身可能行动的小规模模型”，使我们能够“尝试各种替代方案”并“在未来情况出现之前做出反应”。

在随后的几十年里，这个想法不断演变，最显著的是预测处理理论。该理论认为，感知——甚至可能是意识本身——依赖于大脑不断生成关于外部世界的预测，并根据传入的感官数据对其进行更新。

但核心观点依然不变。“关键在于，智能涉及在我们的脑海中构建世界模型，以便我们能够模拟结果并避免代价高昂的错误，”麻省理工学院的认知科学家兼 AI 研究员乔什·特南鲍姆（Josh Tenbaum）说道。“所以，正如克雷克所说，‘我们的想法代替我们死去’。”

不需要天才也能看出为什么这引起了 AI 研究人员的兴趣。LLMs 已证明其能力惊人，而且它们确实能给人一种它们像我们一样理解事物的印象。但语言是对现实的描述，而非现实本身，而它们缺乏直接经验，这意味着即使是最强大的语言模型，在“理解”现实世界中的物理现象时也会感到吃力。

米切尔表示，当被要求对空间概念以及物理世界中可能发生的事情进行推理时，大语言模型（LLMs）的表现通常很差。“如果你曾向这些模型上传过一张地图并就此提问，你就会发现它们在推理方面经常会出现很多问题。”事实上，在2024年的一项研究中，当研究人员使用纽约市出租车行程的逐步导航数据库训练语言模型时，他们发现系统能够提供从一点到另一点的合理路线，但在被要求采取不寻常的绕行方案时，表现得极其糟糕。其原因在于，LLMs 仅拥有行程的描述，因此缺乏那种能让我们想象在不同场景下会发生什么的心理地图。而导航仅仅是一个例子。在任何需要模拟物理世界的场景中——例如安全地装载洗碗机或折叠衣物——LLMs 都会表现出类似的缺陷。

“它们并非“LLMs 缺乏允许我们想象不同场景的心理地图” 8 August 2026 / New Scientist / 31 ---

Google DeepMind 的 Genie 3 能根据文本提示创建用户可探索的世界 为其而设计，这就是为什么人们对世界模型（world models）等替代方案感兴趣，”特南鲍姆（Tenenbaum）说道。

最具有影响力的支持者可能是扬·勒昆（Yann LeCun），他是纽约大学的计算机科学家，直到最近还是 Meta 基础 AI 研究的首席科学家。他在 2022 年的一篇论文中首次提出了他的观点，其核心在于：如果我们想要构建能够真实智能地在现实世界中进行推理、规划并有效行动的系统，我们需要世界模型。“我无法想象在没有预先预测其行为后果能力的情况下，我们能构建出代理系统（agentic systems），”勒昆在 1 月份告诉人工智能论坛 AI House Davos。他认为，关键在于能够通过观察学习世界规则的系统。

这个想法已经流行起来。2025, 12月，勒昆离开 Meta 创立了 AMI Labs，筹集了略高于 10 亿美元 billion 的资金来构建世界模型系统。一年前，加利福尼亚州斯坦福大学的李飞飞（Fei-Fei Li）在获得 2.3亿美元 投资后创办了 World Labs，旨在开发具有“空间智能”的 AI。许多成熟的 AI 公司，最显著的是 Google DeepMind，现在也在以某种形式积极追求世界模型。

资金涌入的主要原因是世界模型在推进机器人技术方面所持的前景（见下文“AI 世界模型有什么用？”）。然而，目前仍处于早期阶段，现有的原型在应用方面较为简单。例如，World Labs 通过 Marble 构建了一个能根据文本提示生成连贯 3D 场景的系统。同样，DeepMind 的 Genie 3 能创建令人信服的交互式虚拟环境——比如“黄昏时的雪山”——你可以在其中移动几分钟，甚至触发如下雨等事件。在这两种情况下，公司都将其系统描述为“世界模型”。但 Marble 实际上是一个 3D 视频生成器，而 Genie 3 是一个视频游戏模拟器，尽管它产生的模拟环境可以让代理在其中以合理的方式行动、观察后果并学习。可以说，一个真正的世界模型应该是存在于代理内部的东西，使其能够预测决策的后果，想象未来，并在采取行动前提前规划——而这正是 DeepMind 通过其 Dreamer 4 系统所努力推进的方向，

AI世界模型有什么用？对于任何好奇AI中世界模型的兴起对我们意味着什么的人来说，研究人员构思的应用场景极具启发性。简而言之，世界模型不太可能产生另一个像ChatGPT那样的时刻——即数百万人突然发现一个可以用在日常生活中的强大工具。其承诺并非更好的聊天机器人，或者任何消费者在短期内能察觉到的东西，而是一代全新的AI系统，其最早的影响可能体现为能够在工厂和仓库中工作的自主机器人，随后可能进入我们的家庭。其逻辑很简单：与其通过在现实中反复尝试并失败来学习（这很危险），或者通过详尽的训练数据学习每种情况的响应（这很昂贵），由AI世界模型控制的机器人可以预测大量场景下会发生什么，并选择最安全且最有效的行动。“机器人技术在改变一切方面具有巨大的潜力，”在旧金山Google DeepMind从事世界模型研究的Danijar Hafner表示。除了机器人，世界模型还被吹捧为加速科学进步的一种方式，特别是在材料科学和药物研发领域，这些领域的挑战在于探索海量的候选分子或原子排列以寻找新型化合物。通过学习底层的物理规律，它们可以在实验室制造或测试任何东西之前，模拟无数种可能性——例如，分子如何与蛋白质相互作用，或者哪些假设的化合物能最好地捕捉二氧化碳。“关键能力是在想象空间中进行规划，因为一旦你拥有了这个抽象的预测器，例如如果你想开发一种药物，你可以说：‘好吧，如果我加入这种化学物质或执行这一系列操作会怎样’，你实际上可以预测会发生什么，而无需在现实世界中操作，”罗德岛州布朗大学的Randall Balestriero说道。“这只是让事情变得更便宜、更快捷。”

世界模型可能有助于人形机器人 32 / New Scientist / 8 August 2026 GEORGE DEEPMIND IMAGE / JAMES GREGG / VIX GETTY IMAGES --- 于2025年9月发布。它学习一个关于环境的内部预测模型，主要基于视频游戏《我的世界》（Minecraft）的数据进行训练，并利用该模型训练一个能够反复“梦见”可能行动之后果的智能体，以改善其行为。“（与Genie 3相比）最大的区别在于Dreamer 4是一个智能体，而不仅仅是一个世界模型，”在旧金山DeepMind领导Dreamer 4团队的Danijar Hafner表示。“它预测行动，并通过迭代自我改进，利用规划和（想象中的）试错来优化这些行动。”

玩游戏

这种方法的威力在 Hafner 及其同事的演示中显而易见：Dreamer 4 能够自行琢磨出在《我的世界》（Minecraft）中如何收集钻石。这是一项复杂的任务，涉及数千种不同的动作——采集资源、制作工具、在地形中导航——而系统在过程中并未被告知如何操作。“系统必须理解其环境并进行泛化，因为每个新剧集都始于一个随机生成的世界，”Hafner 表示。这是一个重要的步骤，尤其是因为这正是能够促成自主机器人的世界模型，例如让机器人能够折叠衣物或装载洗碗机。“我认为这将解决机器人学问题，”Hafner 说。“现在缺失的是执行力：数据、算力和规模化。但我们已经掌握了配方，所以就像语言模型一样，关键在于规模化和细节。”

鉴于 DeepMind 令人印象深刻的往绩，谁会赌它会失败呢？毕竟，其研究人员因在蛋白质折叠方面的工作而获得了诺贝尔奖。但虽然 Dreamer 4 证明了拥有内部世界模型的智能体能够进行推理和规划，它仍留下了一个更深层且尚未解决的问题：这些系统究竟应该学习关于世界的什么内容，或者更具体地说，学习到什么样的细节程度？

这已成为该领域的一个关键分歧点，出现了两种截然不同的方法。许多现有策略试图在“动作条件”下生成预测（正如研究人员所言），即尽可能忠实于训练数据地重建未来的观察结果。但 LeCun 认为这是错误的方法，或者至少不是最好的方法。他的论点基于这样一个事实：人类在心理上模拟世界时并不会关注极大的细节，更不用说逐像素地模拟了。例如，当我们想象一个玻璃杯从桌上掉下来摔得粉碎时，我们并不会预测每一块玻璃碎片、每一滴水珠的位置。LeCun 主张，相反，我们运行的是高度压缩的模型，仅捕捉关键方面。

这就是为什么 主张采取另一种方向，他将其称为联合嵌入预测架构（JEPA）。在这个框架中，世界模型推断出对于推理、规划和行动至关重要的抽象表示。“生成式模型试图重建像素，而 JEPA 在潜空间中学习，并且仅预测有用的内容，”罗德岛州布朗大学的 Randall Balestriero 表示，他曾与 合作开发基于 的世界模型。“它忽略无关细节，仅专注于对智能体至关重要的部分。”

及其投资者似乎在押注 方法将提供一条通往现实世界应用的更快捷路径，这在很大程度上是因为它所需的训练数据更少。到目前为止，该技术最具体的公开演示是一个名为 V- 2 的系统，由 Meta 于 2025 年 6 月发布。V- 通过掩盖视频片段中的某些区域（例如在连续多个帧中遮住一个移动的球）进行训练，并重复预测被隐藏内容的抽象表示而非像素，从而学习一个关于球如何移动的紧凑内部模型。因此， 及其同事证明了他们的系统可以在无需详细重建物理世界的情况下，对其因果结构进行建模。

这并不是说 JEPA 必然比生成式世界模型更好。其支持者当然认为它会更好。“如果没有抽象化，AI 系统将始终局限于狭窄的任务，”Balestriero 表示。但 V-JEPA 2 尚未能清晰地证明其抽象表示是否具有足够的意义，以使一个在开放环境下运行的智能体能够进行推理和规划。“最大的缺失环节在于：我们如何知道这种抽象化实际上是有用的，”Balestriero 说。“这正是目前大量活跃研究的方向，旨在理解：你捕捉到了什么，或者你如何对关于世界的极其丰富且对下游规划有用的信息进行编码。”就 Hafner 而言，他不认为对现实更压缩的表示就更好。“Yann 在很多事情上是对的，➤ 8 August 2026 / New Scientist / 33 --- New Scientist 音频 您可以收听许多文章——请在我们的应用程序中寻找耳机图标 newscientist.com / app 即使他表达这些观点的方式可能比必要的更具争议，而且 JEPA 方法非常有前景，”他说。“但我不认为表示应该是压缩且微小的。最终，你希望学习到强大的表示，而我们在 Dreamer 4 [使用像素训练但在抽象空间中预测] 中所做的工作是非常鲁棒的。”

智能思考

更广泛地说，在现阶段还远不清楚，目前正在开发的世界模型——无论它们是通过高保真地重建世界还是通过推断抽象表示来学习和预测——最终是否足以带我们走向通用人工智能（AGI）。诚然，AGI 同样是一个具有弹性的术语，而关于将世界模型作为实现 路径的主张也存在于一个光谱之中，其中一些最雄心勃勃的观点认为，对物理现实的学习模拟可以成为通用能力推理智能体的基础。例如，DeepMind 在发布 Genie 3 时坚持认为，世界模型是“通往 道路上的关键敲门砖，因为它们使得在无限且丰富的模拟环境课程中训练 AI 智能体成为可能”。事实上，Hafner 认为，“如果你拥有能够表示支配世界之规则的系统，那么你就正在接近类似 的东西，因为这就是理解，这是智能的关键部分”。就 LeCun 而言，他将世界模型视为更广泛系统中的一个至关重要组件，而非全部。

然而，探讨目前开发中的这类世界模型在多大程度上能接近人类智能是值得的，因为这可以揭示可能还需要哪些其他条件。Hafner 表示，最紧迫的需求是时间抽象——也就是说，不仅能够推理近期将发生什么，而且能够推理事情在持续演进过程中的情况。“你不能仅仅在亚秒级、逐帧地模拟一切，因为人类不是那样推理的，”他说，“这是开放的前沿领域之一。”Tenenbaum 则走得更远。他说，人类的世界模型不仅仅是预测引擎，“它们比这丰富得多”。人类的推理依赖于各种形式的因果抽象和假设驱动的推断——“当今 AI 系统的一个局限性是缺乏对其自身认知状态的意识”

我们的心智拥有能够让我们预测接下来会发生什么的模型，更不用说对他人的心智模型了——这些模型共同使我们能够在不同情境下灵活地重新组合知识。“这一切中的一个关键开放性问题是，扩大这些世界模型的规模是否能恢复那种丰富性，而我的观点是，这很可能无法实现，”特南鲍姆（Tenenbaum）说道。米切尔（Mitchell）提出了一个类似但范围略广的观点。“我认为，能够对世界的某些方面拥有某种压缩的、可模拟的表示形式是非常重要的，而这种世界模型的概念可能会带来有用的改进，”她说。“但我认为，智能中可能还有许多其他重要的方面。”米切尔表示，当今AI系统的另一个局限是缺乏元认知——即对其自身认知状态的意识，对其已知和未知之事的意识，以及对其不确定程度的意识。“这能用世界模型解决吗？嗯，显然这取决于什么样的世界模型，”她说。“但我有点担心‘世界模型’这个概念会被用作描述我们现状与‘通用人工智能（AGI）’之间差异的术语。”所有这些都表明，目前构想的世界模型对于实现人类水平的机器智能可能是必要的，但未必充分。事实上，虽然它们在机器人领域，以及可能的科学模拟中，看起来越来越可能变得强大且真正有用，但它们能否复制人类认知中世界模型运作的方式，更不用说更广泛的人类水平智能，这远非板上钉钉的事。“一个巨大的误区是认为智能是一个单一的事物，认为大脑中只有一个单一的世界模型，”特南鲍姆说。“我们实际上拥有的是根据语境、任务和目标运行许多不同模型的能力。”因此，最终，要复制我3岁女儿的大脑在模拟和预测物理世界方面所具备的能力——即便这不能阻止她把水杯从桌上撞下去——是一项不可低估的任务。■

Daniel Cossins 是一位记者，曾任《新科学家》杂志的特写报道负责人 34 / New Scientist / 8 August 2026 --- New Scientist 探索之旅 # 在地球上最佳的观测点观看下一次日全食 由天文学专家引领的科学之旅和游轮之旅，旨在将您带到全食带上 2027 埃及 / 摩洛哥 / 西班牙 2028 澳大利亚 / 新西兰 - 每一次旅程都经过精心规划，以确保靠近日食中心线 - 与领先的天文学家和日食专家同行，他们将为您解释该现象背后的科学 - 与同样充满好奇心的旅行者共同享受这次体验 尽早预订以确保您的名额 探索所有日食之旅和游轮之旅 newscientist.com / tours

--- 特写

CSI：微生物组

每个犯罪现场都包含隐藏的微生物线索，但我们能否利用它们来破案？马修·庞斯福德（Matthew Ponsford）对此提出了疑问。

2015年，两名入侵者闯入了佛罗里达州好莱坞的何塞·洛佩兹（Jose Lopez）家中并偷走了一台电视。但这并非一起典型的盗窃案。这些“窃贼”其实是当地警长办公室的副警长，而洛佩兹——诺瓦东南大学（Nova Southeastern University）的一名生物学家——参与了整个计划。这是一个研究项目，旨在探讨警方是否能通过犯罪分子在现场留下的微生物来识别其身份。在分析了表面留下的微生物痕迹并分离出细菌DNA后，研究人员能够将它们与这些副警长匹配。但当他们尝试从一个包含1万,000个微生物组样本的数据库中挑选出入侵者时，这些特征证明其独特性不足。他们在2020年的一篇论文中得出结论，该技术“不能作为刑事调查中可靠的痕迹证据标准”。我们可以将其视为存在“合理怀疑”的理由。

法医学专家长期以来一直认为，我们身体内部和表面的数十亿个细菌、病毒和真菌可能成为宝贵的证据来源。问题在于如何证明“微生物组法医学”在实际调查中足够可靠。然而，近年来研究人员改进了技术，与警方合作以更好地确定这些方法能告诉我们什么，以及不能告诉我们什么。他们还开发了一种分析死后微生物组以估算死亡时间的方法——这种方法目前正被用于探索一起意大利极具争议的谋杀案。

“这是目前法医学领域的一个巨大前沿，”夏威夷檀香山查米纳德大学（Chaminade University of Honolulu）的大卫·卡特（David Carter）表示。问题在于，这些技术中哪些将被证明有用，以及它们产生的证据是否能在法庭上站得住脚。

不难理解为什么法医学专家将人体微生物组视为一个尚未开发的资源。“想象一下，如果我告诉你有一种物理证据保证会与每一个死亡现场相关联——无论地点在哪里，时间在何时，”卡特说道，他暗示了许多指纹和DNA无法检索的案例。在这种情况下，我们不断脱落到他人身上以及我们接触到的任何

照片：M. Ponsford 36 / New Scientist / 8 August 2026 ---

表面上的微生物可能会提供线索，因为例如我们皮肤上的生态系统在人与人之间存在巨大差异。

法医微生物学领域实际上直到20世纪90年代才出现，当时测序技术的进步改变了分析微生物DNA以分离特征细菌物种的可能性。随着测序速度的加快和成本的降低，研究人员开始测试这样一个想法：我们携带的高度个体化的微生物混合物——即我们的微生物“指纹”——可能在识别人员方面有用。例如，在2010年发表的一项研究中，现就职于加州大学圣迭戈分校的罗布·奈特（Rob Knight）及其同事表明，沉积在个人键盘或鼠标上的细菌组合可以从270人的数据库中识别出所有者。与此同时，其他研究表明，人们在自己的家中会留下独特的微生物签名。

这看起来很有前景。然而，当研究人员在更真实的条件下（包括那些模拟盗窃案中）测试“接触微生物组”方法时，它远未达到法医证据所要求的可靠性。

这些结果促使当时芝加哥大学微生物组中心主任、且领导了入室盗窃研究及其他相关研究的杰克·吉尔伯特（Jack Gilbert），在2018年警告美国司法部（DOJ）不要采用这项技术。吉尔伯特表示：“我明确指出，我不认为这项技术能以足够的严谨性来识别个人，从而断定那就是他们而不是其他人。我无法毫无疑问地证明，我发现的细菌不是来自另一个人类。”

不过，这次挫折并未终结人们对法医微生物学的兴趣。此后，其他研究人员不仅在探索更好的测序技术和更大的数据集能在多大程度上产生影响，还在研究其他形式的微生物组分析是否能在刑事调查中发挥作用。英国兰开夏大学的法医科学家诺埃米·普罗乔皮奥（Noemi Procopio）正在研究其中一种可能性。她认为，接触微生物组技术可能已经拥有某种“调查情报”，在缺乏传统法医证据的案件中为警方提供关键线索。

她目前的研究项目始于安迪·尼克森（Andy Nixon）的挫败感，尼克森是英国西米德兰兹警局的一位资深犯罪现场专家，他经常面对一些在他看来存在“法医真空”的谋杀案。尼克森说，许多最令人头疼的案件涉及勒杀，因为“颈部过于柔软且具有弹性，无法保留传统的指纹，而单纯的接触很少能留下可用的DNA片段——即使攻击者使用了如此极端的暴力”。

调查工具箱

为了验证微生物痕迹是否能提供帮助，Procopio和Nixon在2025年从西米德兰兹郡警察局招募了100名新警员，以测试其接触微生物组被追溯到本人的可靠程度。参与者将手指按在醋酸纤维片上，这些纤维片被放置在人们可以走过的房间里数日或数周，以模拟现实世界中微生物指纹可能存在的情况。

起初，Procopio及其同事专注于稀有细菌物种，这些物种有时能识别个体，但通常在48小时内通过测序就变得无法检测，这很可能是因为它们被来自环境的其他物种所掩盖。随后，研究人员分析了样本中微生物群落的整体结构，以观察其是否能提供更可靠的特征。“也许我们拥有相同的细菌，但我拥有某种细菌更多，另一种更少，从而形成某种独特性，”Procopio说道。在尚未发表的研究中，他们发现这种方法使他们能够以更高的准确率将样本与个体匹配，且这种信号在一个月后依然存在。

至关重要的是，Procopio并未声称此类技术将取代DNA分析，或足以单独将嫌疑人定位在犯罪现场。相反，她将这些技术视为调查工具箱中的“补充”手段。事实上，她同意Gilbert的观点，即“在现阶段，要达到通常与人类DNA鉴定相关的那种统计权重将非常困难”。

8 August 2026 / New Scientist / 37 ---

但在调查人员遭遇法医学空白的情况下，如果微生物群落的整体结构与从已知嫌疑人身上采集的样本相匹配，他们可能会获得线索。Knight表示，这反映了法医学证据在实践中的常见用法。最常见的法医学发现并非某种结论性的鉴定方法，而是能够缩小嫌疑人范围的“类别证据”。

Knight提到了一起1982年发生在佐治亚州亚特兰大的儿童谋杀案，调查人员在几名谋杀受害者身上以及嫌疑人韦恩·威廉姆斯（Wayne Williams）拥有一块不寻常的地毯中，都发现了微观纤维的混合物——黄绿色尼龙和紫色醋酸纤维。根据销售记录，他们向陪审团解释了在嫌疑人家中发现此类地毯的极低概率：1 in 7792。

“单根纤维并不能得出结论，但综合起来则非常有说服力，”Knight说，“在这种语境下对待微生物是合理的。”微生物痕迹可能起到类似的作用。

指纹和DNA并不总是出现在犯罪现场

“微生物组分析可能揭示关于未知凶手的线索”

“一项技术不需要完美才能发挥作用，”Knight 表示，“关键在于叠加多个类别特征，使错误匹配的概率趋近于零。”

我们随身携带的微生物群落也可能指向调查人员此前未曾关注的方向。这是因为我们的微生物组在很大程度上是由我们的生活方式以及一生中与他人的互动所塑造的，从居住地、相处的人到饮食习惯皆然。例如，在 2023 年，Procopio 及其同事证明，基于口腔微生物组，微生物组分析可以区分意大利两个相邻地区——伦巴第（Lombardy）和皮埃蒙特（Piedmont）的居民。

“这绝对是有用的，”Procopio 说。这里的挑战在于参考数据。目前，我们还没有足够大规模的多元个体微生物组数据集来得出有用的结论。与 DNA 一样，庞大的数据库能带来更深刻的洞察。但 Procopio 认为，我们最终可以使用微生物组分析来揭示关于未知凶手生活方式的线索，例如——他们是在办公室还是在农场工作，或者他们是否有特定的健康问题，甚至是否养宠物。

话虽如此，尽管利用微生物识别犯罪现场作案者的尝试充满吸引力，但微生物组在法医学中最有前景的用途是揭示某人死亡的时间——以及在某些情况下的死亡方式。Procopio 是一组研究人员中的一员，他们试图利用尸体分解过程中出现和消失的细菌来开发“微生物钟”，以便在发现尸体数日甚至数周后，仍能对死亡时间提供可靠的估计。

近期意大利一起最具争议的谋杀案比任何案例都能更好地说明这种潜力：利莉亚娜·雷西诺维奇（Liliana Resinovich）案。2022 年 1 月，在 63 岁的雷西诺维奇被报告失踪近一个月后，她的尸体在的里雅斯特（Trieste）一家前精神病院的院内被发现。发现时，她的头部和身体被塑料袋包裹，但法医病理学家没有发现明显的受攻击迹象；他们的第一份报告结论是，她在之前的 48 小时内死亡，可能是自杀。

但随后，在雷西诺维奇家属的抗议下，她的尸体被重新检查。这次新的法医评估不仅得出结论认为，由于在面部和头部发现了创伤迹象，她是死于窒息，而且雷西诺维奇是在失踪前后被杀的——引用的证据包括她失踪当天早晨最后一餐尚未消化的食物残渣。

“有人说她死了 48 小时，然后下一个法医专家说她已经死了 23 天，”Procopio 说，“这是一个巨大的差距。”

至关重要地，这给真正发生了什么留下了疑问：雷西诺维奇是在尸体被发现前两天死亡的，如第一份报告所言？还是她在失踪前后被杀，并被保存在某个寒冷的地方，甚至可能是被冷冻，直到她的尸体被

38 / New Scientist / 2026 年 8 月 8 日 ---

我们在环境中都会留下独特的细菌签名，通常来自我们的手指转移到她被发现的地方——这是一个为了解释为什么一具在户外放置了三周的尸体看起来保存得异常完好而提出的假设？

尽管侦探剧中的情节可能有所暗示，但传统的法医学工具只有在尸体被迅速发现时才能精准判定死亡时间。Procopio表示，在最初的24小时内，法医病理学家通常可以使用直肠温度计的读数来计算一个可靠的“死后间隔”；而在最初的几天里，肤色变化、尸僵以及在尸体孔口产卵的昆虫线索则能提供帮助。但她补充道，随着尸体进入更高级的腐败状态，不确定性会增加，尤其是在户外的情况下。Procopio指出，这正是微生物钟旨在解决的那类不确定性。在过去的十年中，针对美国各地“尸体农场”中人类遗骸的研究表明，分解过程以一种特定的方式进行：不同的细菌以可预测的模式出现、繁盛并衰减。通过追踪这些微生物的变化，研究人员可以估算尸体分解的时间，从而有可能挖掘出关于某人何时以及如何死亡的关键线索。在Resinovich案中，在发现尸体前后采集的口腔拭子保存了其口腔微生物组的快照。如果她的尸体在被发现之前被冷冻或保存在寒冷的地方，那么该样本中的微生物群落应该包含与冻融循环一致的改变证据。Knight表示：“这会留下可检测的特征。”Procopio补充说，其他线索可能来自昆虫，因为3月份发表的一项研究表明，昆虫在不同时间将某些分解微生物引入尸体，这取决于尸体是从死亡那一刻起就暴露在户外，还是从室内被移至室外。在Procopio看来，微生物组可以解决“冷冻机假设”与“自杀假设”之间的不确定性。她说，到目前为止所有的证据“显然是指向一个方向而非另一个方向的”。但为了让证据在法庭上具有可采信性，法官需要有令人信服的证明，证明微生物钟是可靠的。这就是为什么Procopio及其同事现在正在密歇根州北部的一个尸体农场重建这两种情景，那里的气候条件与意大利北部相似。在那里，四具捐赠的尸体被保存在不同的条件下，使研究人员能够追踪其微生物组随时间的变化，旨在向负责Resinovich案的公诉人提交一份清晰且可采信的结论。至于微生物组法医学是否能产生导致刑事定罪的证据这一更广泛的问题，答案取决于具体应用。在大多数国家，法官充当“把关人”，有权决定是否允许将一种新型科学证据引入法庭，这通常遵循类似于美国自1993年起制定的指南。在后来被称为“道伯特标准”（Daubert standard）的准则下，法官在将一项技术接纳为证据之前，会考虑该技术是否经过独立测试和同行评审，以及是否具有已知的误差率。Carter表示，在所有这些方面，用于估算死亡时间的微生物钟似乎都已经准备好在法庭上使用了。由科罗拉多州立大学的Jessica Metcalf领导的研究表明，如果尸体在10天内被发现，则可以做出误差率为三天的可靠死后间隔判断，不过死亡时间越长，误差就越大。Carter说：“我觉得它符合这些标准，但最终决定权确实在法官手中。”

利用微生物组分析来识别嫌疑人的前景则更为不确定。若要作为识别嫌疑人时 DNA 比对的替代方案，它必须达到同样严苛的证据标准——即误报率大约为百万分之一。这里的难点在于，皮肤微生物组仍然是一个如此动态且不断变化的生态系统，以至于吉尔伯特怀疑它在短期内不会有用。“微生物组包含 500 个物种，可能有 5000 到 6000 个菌株，它们的丰度在不断波动，不断经历遗传进化，比例代表性在不断转移，并且在人与人之间不断交换，”吉尔伯特说道。“你面对的是生物学，而生物学极其复杂。”其他人则更为乐观。奈特认为，自 2018 年以来，测序技术在关键方面已经取得了进步。他说，更先进的方法可以检测出微生物之间不那么明显的遗传差异，从而揭示旧方法无法发现的“个性化且持久的菌株”。“我不会放弃痕迹证据这个课题，”奈特说。“虽然具有挑战性，但技术正在迅速进步。”

Matthew Ponsford 是一位常驻伦敦的记者 2026年8月8日 / 《新科学家》 39 --- 特写 访谈

“悲伤发作本质上是一次全人类体验式的攻击” 几年前，心理学家谢尔曼·李（Sherman Lee）经历了一次“悲伤发作”（grief attack），并开始对其进行研究。他告诉瓦妮莎·尼罗德（Vanessa Nirode），这种发作是多么强大且令人心力交瘁。

40 / New Scientist / 8 August 2026

---

有一次，谢尔曼·李在开车时突然剧烈抽泣，以至于失去了对肌肉的控制，为了防止车祸，他不得不将车停在路边。他的全身都被对一个月前去世的心爱宠物的悲痛所淹没。他以前经历过恐慌发作，但这似乎有所不同；他不仅是呼吸困难，而且充满了强烈的绝望。现在回想起来，李说这次经历感觉像是一次“悲伤发作”，这个术语是在几十年前创造的，并在心理学界被广泛认可，但尚未经过严谨的研究。去年，作为弗吉尼亚州克里斯托弗纽波特大学（Christopher Newport University）的心理学家及悲伤研究专家，李共同发表了一项针对近 250 名经历过悲伤发作的人员的研究，以更好地定义这一现象。

李的研究契合了人们对悲伤作为一种复杂且非线性体验的日益深入的理解。“悲伤的五个阶段”这一概念早已被推翻，心理学家现在强调，悲伤可能持续数年，影响身心，并渗透到生活的方方面面。它既令人极度痛苦，又是最深刻且本质上属于人类的体验之一。在接受《新科学家》采访时，李解释了他关于悲伤发作的发现，以及理解悲伤之 myriad（无数）表达方式的重要性。

瓦妮莎·尼罗德：悲伤有哪些不同的表现方式？

谢尔曼·李：只要你提到一个人生活的任何一个方面，我都能向你展示悲伤是如何在其中表达的。在生物学上，悲伤体现在睡眠行为或饮食紊乱中。由于压力，处于悲伤中的人免疫力可能会下降。如果一对夫妇在一起的时间非常长，那么遗孀或鳏夫很有可能在不久之后随之而去。在心理学上，悲伤会产生强烈的情绪，如渴望、悲伤或愤怒。人们会沉溺于对失去之人的思念，尤其是如果那是一次悲剧性的死亡。很多人会出现幻觉。在一项针对 1万,000 多名经历过悲伤的人员的大规模研究中，50% 的参与者报告感觉到某种“存在感”；三分之一的人感知到听到了对方的声音。在社交方面，人们可能会避开任何能触发对逝者记忆的事物。他们的生活圈变得非常狭小。他们可能会避免从事过去与逝者共同参与的活动。或者，他们可能会通过某种形式的药物滥用来应对。在精神层面，你可能会对上帝感到愤怒，不再相信你曾经相信的事情——这是一种巨大的存在主义危机。

“悲伤发作”这个概念对我们中的许多人来说可能并不熟悉，但我相信它已经存在一段时间了。

特蕾莎·兰多（Therese Rando）是悲伤治疗领域的先驱。她最早在 1991 年探讨了悲伤发作这一主题（尽管她当时将其称为“悲伤的突然激增”），并基于临床观察撰写了大量相关文章。相比之下，我的同事和我则利用来自大量人群的数据，识别并科学地验证了这些症状。

那么，究竟什么是悲伤发作（grief attack）？ 我们能够识别出四个主要特征：恐慌（表现为头晕、呼吸短促和恐惧）、渴望、绝望以及思维混乱。我们还创建了一个用于识别和评估这些症状的工具。[悲伤发作]与“标准”的悲伤不同，因为它们既包含身体症状，也包含情感症状，且其强度通常是突然袭来的。我们目前还不知道它们在频率上的普遍程度，但在我读过的许多书籍和文章中，它们经常被提及，尽管名称各异，如“阵痛”（pangs）或“痉挛”（spasms）。我认为这一切都向我们表明，这种情况相对常见。最重要的一点是，这是一种真实的病症，因此可以进行科学研究，并在必要时进行治疗。

悲伤发作与恐慌发作有何不同，为什么要做出区分？ 从基础科学的角度来看，它将两个通常被分开研究的系统结合在一起：恐惧系统和依恋系统。悲伤发作看起来像是这两个系统在同一时间发生的独特风暴——2026年8月8日 / New Scientist / 41 ——出汗和颤抖是恐惧系统的表现，而渴望的想法则是依恋系统的一部分。当你经历悲伤发作时，除了呼吸短促和头晕等典型的恐慌发作症状外，你还会经历急性悲伤症状，如渴望、无法抑制的哭泣以及对逝者的思念。这两者的结合强度使其成为一类独特的人类体验。

我们是否知道为什么有些人会经历这种情况而其他人不会？ 我假设有两个主要因素可能对此有所影响。首先，一个人[如果具有]焦虑敏感性（anxiety sensitivity），可能会经历悲伤发作，这种敏感性是指担心心跳加速等焦虑症状会对身体造成伤害。例如，如果你的心跳加快或无法呼吸，有些人会将其解读为对生命的威胁，因此他们会感到非常恐惧和恐慌。我恰好属于这一光谱。我们知道，具有焦虑敏感性的人被诊断为恐慌障碍的可能性高出五倍，而由于悲伤发作包含恐慌发作的成分，因此可以推断他们也更有可能经历悲伤发作。第二个因素是压力的累积，这可能会触发恐慌发作并使悲伤症状恶化。如果一个人的生活中存在复合压力，他们可能更有可能出现悲伤发作。

人们被悲伤和悲伤发作深度影响的时间有多长？ 我认为当我们说“深度影响”时，大多数人的意思是悲伤导致功能失调或极端痛苦的时间有多长。悲伤研究人员认为大约是六个月。最新的《精神疾病诊断与统计手册》（DSM）则给出了整整一年。在那之后，我们开始讨论延长悲伤障碍（prolonged grief disorder）。但在我看来，“深度影响”意味着在根本上改变了你。我认为这是终身的。以C.S.路易斯为例，他写了一整本书记录失去妻子的经历。许多人能够正常生活、欢笑并拥有充实的生活，以至于当你与他们交谈时，你不会知道他们失去了重要的东西，但每当他们绘画或写作时，那种感觉就在那里。[悲伤发作]可能发生在失去亲人后生命历程中的任何阶段，尽管它们在悲伤过程的早期似乎[更为]频繁。

“那些允许悲伤发作来去的人，比那些试图分散注意力的人情况更好”

人们在面对悲痛发作时，有什么方法可以管理这种体验吗？ 我们询问了人们是如何应对的，尽管我们没有直接进行任何实验。我们的研究表明，那些允许这种体验自然来去的人，比那些试图分散注意力或专注于某项活动的人情况更好。这就是正念教给我们的：不要评判它，不要对抗它，只需允许这种体验发生，因为它在几分钟后就会消失。

关于悲伤如何影响大脑，以及在悲伤发作期间我们的大脑中发生了什么，我们目前了解多少？ 认知测试表明，悲伤的体验越复杂、越剧烈，我们在记忆力和注意力方面看到的损害和紊乱就越多。一些研究显示，在极度悲伤的人群中，大脑体积较小。此外，慢性压力会导致神经元丢失，这可能会体现在大脑体积上。它还会影响处理奖励和成瘾的大脑区域，以及与身体疼痛相关的区域。42 / New Scientist / 8 August 2026 --- 失去挚爱可能会导致突然的、类似于恐慌发作的片段。悲伤发作可能包括强烈的渴望感和绝望感。在悲伤发作期间，大脑的多个区域会同时激活。由于恐惧因素，杏仁核会参与其中，类似于“战斗或逃跑”反应。导水管周围灰质 [大脑中调节疼痛调制和“逃跑或冻结”反应的区域] 会启动，因为你渴望那个人回来。大脑奖励系统的伏隔核会由于一种渴求方式而寻找那个人，从而被触发。然后是前扣带回，即大脑的疼痛中心，疼痛体验与情感在此汇合。悲伤发作本质上是一次全人类体验的冲击。您如何判断悲伤或悲伤发作是变得令人担忧，还是一个自然过程？ 我认为我们所说的“自然”是指好的：什么是良好或健康的哀悼？这在某种奇怪的程度上假设了“好”必须是无痛的。你爱一个人，而这种爱的自然结果就是因断联而产生的痛苦。你必须面对这种痛苦直到它消退。虽然悲伤发作的干扰程度足以令人恐惧，但我不会说它们令人担忧或具有直接的身体危险。在我的发作之后，我被吓到了，以至于那天剩下的时间我不想开车，但我仍有足够的控制力将车停在路边。我认为 [我们研究中的] 大多数人也是这样。在心理上，它们可能让你觉得自己在失去控制，这很可怕，但并不一定危险。与此同时，心理事件具有干扰性且是真实的。我们不应忽视它们，人们在需要时应该寻求支持。每当你发现悲伤过于沉重而无法处理，或者当它损害了你的功能能力时，你应该寻求专业帮助。听起来我们中只有相对少数的人会经历悲伤发作，但几乎我们所有人都会在生命中的某个时刻感受到悲伤。对于处于这种情况的人，您有什么建议？ 对很多人来说，结清（closure）很重要：仪式、葬礼、能够说出最后的话。在试图理清头绪并获得视角时，与他人谈论失去是有帮助的。有证据表明，身体活动、锻炼和运动对于那些在哀悼失去之痛的人很有帮助。除了对抗孤独和抑郁，锻炼和运动还可以帮助解决睡眠问题，尤其是如果你能让自己疲惫到足以压制过度思考的大脑。此外，如果你在做一项运动，你就不会在家里看照片；你会获得暂时的分心。我们都需要从痛苦中休息。当你决定寻求专业帮助时，悲伤治疗是为你个人量身定制的。因为悲伤是一件大事。它绝不仅仅是一件事。许多人可能会建议认知行为疗法，它处理的是给你带来麻烦的想法和行为。假设困扰你的是悲伤发作，那么我会建议基于接纳的疗法，其目的是教会你让它们发生并接纳它们，从而建立心理灵活性以及一种让自己立足于当下的方式。

所有这些都是极佳的疗法和方法，但很难为[悲伤]提供一个万能药。它包含许多方面。我建议尝试找出在这个大问题之中，你具体遇到了哪些问题。■

Vanessa Nirode 是一位撰写健康与福祉内容的记者 8 August 2026 / New Scientist / 43 --- 区别在于 NEVILLE JOHNSON

无需杂乱，无需烦琐，最重要的是无需施工，即可改造您现有的楼梯。将您的楼梯翻新为一件精美的艺术品，比您想象的要简单得多，成本也更低。仅需1-2天且无需施工，我们就能用一座惊艳的手工打造楼梯来改变您家的核心区域。从让空间感觉更轻盈、更开阔的玻璃设计，到提升楼梯整体外观的优雅钢制细节，每项设计都根据您的家居、风格和预算量身定制。 ✓ 英国楼梯翻新市场的领导者 ✓ 由专家在48小时内安装完成 ✓ 采用优质材料定制生产 ✓ 10年质保 ✓ 提供金融方案 申请手册 0161 873 8333 / NEVILLEJOHNSON.CO.UK --- # 封底页 益智游戏 ## 快速填字游戏 #214 由 Richard Smyth 出题 涂鸦区 下周预告：隐晦填字游戏 ### 横向 1 自动机 (6) 4 医学成像程序 (1,1,4) 9 概念，发明 (4) 10 安非他命商品名 (10) 11 喉部的淋巴器官 (6) 12 Lynn ___，美国进化生物学家 (8) 13 痈簇 (9) 16 导致痤疮的皮肤状况 (4) 17 圆圈 (4) 18 100 万 毫米 (9) 21 由 Jacques Cousteau 共同开发的潜水品牌 (4-4) 22 Ag (6) 25 大约始于 5.4 亿 年前的地质年代 (10) 26 可接近的 (4) 27 流水的现象 (6) 28 三角函数中的倒数函数 (6) ### 纵向 1 与 Guglielmo Marconi 相关的广播技术 (5) 2 铜 (Cu) 和锌 (Zn) 的合金 (5) 3 细胞蛋白 (7) 5 推测或假设 (6) 6 血细胞，例如 (9) 7 系统中，输出与输入不成比例的 (9) 8 智人 (Homo sapiens) 所属的分类群 (6,7) 14 平坦的 (3,6) 15 例如，一种病毒、毒素或细菌？ (9) 19 中耳的小骨 (7) 20 皱缩的 (6) 23 多巴胺前体 (1-4) 24 体外分泌物 (5) ## 快速问答 #364 由 Chelsea Whyte 出题 1 Hong Wang 成为历史上第三位获得数学界最高奖项——菲尔兹奖的女性。她是因为解决了什么问题而获奖的？ 2 最近发现宇宙中的哪种成分可能并不存在？ 3 虎鲸被观察到冲撞哪种海洋动物，导致其炸裂成数千个碎片？ 4 一家农业初创公司开发了一种提高大豆、辣椒和大麻产量的方法。它使用什么来刺激植物生长？ 5 澳大利亚 Cloggs 洞穴中的什么发现表明，GunaiKumai 人在那里进行复杂的仪式已经超过 1000 代？ ## 块宇宙 (Block Universe) 由 Puzzlr 创建 目标是将网格划分为不重叠的矩形。每个矩形必须恰好包含一个数字，且其面积等于该数字。例如，数字 6 应该有一个覆盖六个方格的矩形。 本周益智游戏的答案见第 47 页 我们的游戏现在可以在线玩 newscientist.com / games 2026年8月8日 / New Scientist / 45 --- # 封底页 几乎是最后一言

分支延伸

我的家谱在每一个过去的祖辈世代都会翻倍，向过去分支延伸。在哪个时间点，它会因为历史人口规模较小而开始减少？ 希拉里·肖 (Hillary Shaw) 英国，什罗普郡，纽波特

全球人口在 1930, 10 亿美元 年约为 20 亿美元，1820, 5 亿美元 年为 10 亿美元，1640 年为 5 亿美元，900. 年为 2.5 亿美元。与此同时，如果你的父母出生于 1950, 年，且 10 次翻倍等于 1000 个祖先，那么理论上你在 1700 年会有 2000 个祖先（假设平均母亲年龄为 25 岁），1450 年有 200 万 个，1200. 年有 20 亿美元 个。因此，这两个序列在大约 1300. 年时相交。

然而，你的家谱数量稳定下来的时间要晚得多，且规模更小，因为在铁路出现之前，许多人生活在孤立的社区中，并与远亲结婚。一个极端的例子是，苏格兰的圣基尔达群岛在 1800 年左右开始大规模移民之前，几个世纪以来一直维持在 100 人左右——这大约是避免过度近亲繁殖和遗传问题的最低人数。在英国本土，在铁路出现之前，有巡回的“在 1700 年代，很有可能你的大多数祖先都来自一个只有几 100. 人的村庄。在那之前，几乎全部是近亲繁殖”商人、法官、传教士、士兵和乞丐，但人们通常会在相似的社会阶层、宗教等范围内结婚。在城市，尤其是港口城市，你可能会娶 / 嫁给一个外来者，但铁路之前的村庄几乎像圣基尔达一样孤立，人口规模也相似。大约在 1650, 年，由于道路糟糕不堪，往返于约克和伦敦之间需要七天时间，而那些道路在当时简直就是泥潭

ARTESIA PETT LEE / BRIAN PHILLIPS

Tom Gauld 为《新科学家》创作

以及跑步（“耐力跑与人类的进化”，2004）。对于步行，曲线呈U形，在速度约为 1.3 米 / 秒时，每公斤体重每公里耗氧量最低，约为 165 毫升 O₂。跑步（包括慢跑）是一条直线，耗氧量从 2 米 / 秒时的约 210 毫升 O₂ / kg / km 下降到 6 米 / 秒时的约 200 毫升。这意味着，如果你想在一段给定的、相当长的时间内走完一定的距离，且希望耗氧量尽可能低，那么你应该步行。然而，当分配的时间少于某个特定值（该值对应于稍快的步行速度）时，你应该在尽可能快地跑步和以该稍快步行速度行走之间交替进行。但请注意，这里讨论的是总耗氧量，而非高于正常基础代谢率的额外耗氧量。为了使额外耗氧量最小化，“我的理论是，因为狗不会攀爬且没有空中捕食者，它们往往没有太强的‘上方’概念”，其最佳速度要比总耗氧量的最佳步行速度更慢。

听到了吗？

为什么低空飞行的飞机或直升机从头顶经过时，狗不抬头看？我从未见过狗这样做。（续）

Dyane Silvester 英国坎布里亚郡，安赛德

作为一名曾经的狗主人，以及现在的常驻宠物看护员，我的理论是，因为狗不会攀爬，且作为一个物种，它们没有空中捕食者，所以它们往往没有太强的“上方”概念（除非是向上倾斜的地面）。如果你从高处掉落一样东西，大多数狗会在地板上寻找，而不会意识到物体和地板之间有一张桌子。不过，我曾养过一只阿尔萨斯犬，如果她在花园里而我出现在楼上的窗户前，她能立即定位并到达我所在的房间，尽管她从未能从我们家的任何楼上窗户向外看。她还会注意到并观察鸟类和低空飞行的飞机。但我确实曾难住过她一次：我爬出天窗并出现在屋顶上。她一直没搞明白那个！

David Lewis 英国伦敦

虽然不常见，但我一个朋友去世的西班牙猎犬会看向公园上空飞行的飞机。然后她会对着它们吠叫并追逐。幸运的是，她从未抓到过任何一架！

C. H. Johnson 英国德文郡，科利顿

因为它们知道飞机和直升机不会带来食物或任何狗狗零食。 ■

答案

纵向 1 Radio, 2 Brass, 3 Tubulin, 5 Theory, 6 Corpuscle, 7 Nonlinear, 8 Animal kingdom, 14 All square, 15 Biohazard, 19 Ossicle, 20 Rugose, 23 L-dopa, 24 Egest 27 Eddies, 28 Secant 25 Palaeozoic, 26 Open, 21 Aqua-Lung, 22 Silver, 16 Acne, 17 Loop, 18 Kilometre, 12 Margulis, 13 Carbuncle, 9 Idea, 10 Benzedrine, 11 Tonsil, 横向 1 Robots, 4 CT scan, #214 快速填字游戏 5 Ash from burnt grass 4 Sound waves 3 Sunfish 2 Dark energy 1 The Kakeya conjecture 快速测验 #364

答案 块状宇宙 (Block Universe)

2026年8月8日 / New Scientist / 47

---

封底页

反馈

气球与双倍气球

本期的“反馈”栏目是一次时间旅行练习——因为每一项内容都来自很久以前，而我们不知为何之前遗漏了。我们会感到愧疚，但现在是暑假（至少在北半球是），许多读者可能得一边盯着调皮的孩子，所以我们认为这样勉强可行。

那么，请将思绪回溯到 2008 年。显然，那是第 20 届加拿大计算几何会议在蒙特利尔举行的年份。这次会议展示了大量的研究成果，但记者马修·斯帕克斯（Matthew Sparkes）最近偶然发现了一项由埃里克·德梅恩（Erik Demaine）、马丁·德梅恩（Martin Demaine）和维·哈特（Vi Hart）进行的理论研究。该研究提出了一套“关于气球扭曲结构的通用数学和算法理论”。如果你没听懂，请允许我们澄清一下：这是关于气球动物的几何学。

作者们开发了一个数学框架，使他们能够计算出生成特定形状需要多少个气球，无论是一个狗形（1 个），还是像正十二面体这样复杂的多面体（10 个）。这些自然被称作“柏拉图气球”。在整篇论文中，作者使用“bloon”一词来指代一个等同于气球的数学实体。“反馈”栏目起初不明白为什么需要这个新词，直到我们意识到，这样他们就可以将长度为通常两倍的气球称为“doubloon”（双倍气球）。

出于好奇，想知道作者们之后在做什么，“反馈”栏目进行了搜索。哈特成为了 YouTube 上的科学普及者，随后成为了微软的一名高级经理。与此同时，马丁和埃里克是父子研究员兼艺术家，两人都痴迷于折叠数学。埃里克是麻省理工学院的教授，经常出现在《新科学家》杂志中。近期的论文包括《圆锥体和凸多面体的重叠展开》以及《俄罗斯方块——仅用一种形状就很难实现》。请多给我们提供一些埃里克的作品。

Twisteddoodles

有适合“反馈”栏目的故事吗？

请发送至 feedback@newscientist.com 或邮寄至 New Scientist, 9 Derry Street, London, W8 5HY。邮寄项的审理将会延迟。

测试时刻

继续讨论“迟到”这个话题，读者德里克·隆（Derek Long）指出了一篇 2018 年的论文，题为《使用降落伞预防从飞机跳下时死亡及重大创伤：随机对照试验》。该文报道了一项名为“广泛持有的缺乏治疗均势信念损害随机试验参与度（PARACHUTE）”的研究。（顺便说一句：这是一个真正出色的科学缩写。我们好久没见到这种缩写了。）文中提出的问题在于证据。 “降落伞被常规用于预防从飞机跳下的人员死亡或重大创伤性损伤。然而，支持降落伞功效的证据十分薄弱，且关于其使用的指南建议主要基于生物学合理性和专家意见。” 如果读者还没反应过来：这篇论文是讽刺性的，旨在嘲讽这样一种观点——即便是最符合常识的干预措施也应该经过完整的随机对照试验测试。正如作者所写：“在 2003 年发表于《英国医学杂志》（BMJ）的一项开创性工作中，史密斯（Smith）和佩尔（Pell）对评估重力挑战期间降落伞功效的随机临床试验进行了系统搜索，结果没有发现任何已发表的研究。” “反馈”栏目实际上在 2004 年就读到过那篇综述。作者指出他们“无法确定任何关于降落伞干预的随机对照试验”，并建议“如果循证医学最激进的倡导者们组织并参与一项关于降落伞的双盲、随机、安慰剂对照、交叉试验，每个人可能都会从中受益”。哎呀。十五年后，他们确实做了一次试验。结果显示，“使用降落伞并未显著降低死亡或重大损伤”，但有一个微小的限制条件，即“该试验仅能招募在地面上的小型静止飞机上跳下的参与者”。一张参与者在停在地面上的飞机上不带降落伞跳下的照片生动地说明了这一点。正如作者所言：“我们的发现应该让那些主张在娱乐或军事环境下从飞机跳下时常规使用降落伞的专家们暂时停下来思考一下。”

机智的兹维基

有些概念无法追溯到其起源。其中之一就是“球形牛”笑话，它说明了在处理复杂现实情况时，使用高度简化和抽象模型所面临的困难。故事涉及一名物理学家，他想出了提高奶牛场生产力的方法，但其方案仅适用于真空中的球形牛。似乎没人知道是谁想出了这个点子：最早的参考文献将其定位于 20 世纪 60 年代，但起源故事都很模糊，而且有一篇参考文献把牛换成了鸡。然而，另一个基于球体的物理学家笑话似乎更早。在事情发生几十年后，“反馈”栏目刚刚得知弗里茨·兹维基（Fritz Zwicky）使用过的一种侮辱语，兹维基在 20 世纪 30 年代提供了关于暗物质最早的一些证据。作为一名著名的古怪之人，兹维基喜欢将他讨厌的人称为“球形混蛋（spherical bastards）”。这样的人无论从哪个角度看都是个混蛋。“反馈”栏目今后将采用这个词。但我们也想知道“球形混蛋”、鸡和牛的词源（如果这个词用在这里合适的话）。欢迎提供任何见解。 ■ 48 / New Scientist / 2026 年 8 月 8 日

---

让夏季娱乐变得毫不费力

购买任何配备 CUBE 且由 Quooker 工程师安装的 Quooker 系统即可节省 £401

优惠将于 2026 年 8 月 31 日结束。探索 如何改变您的夏天。只需一个水龙头，即可享受即时的 100℃ 沸水、冰水和气泡水。 水龙头可在两小时内安装在您现有的厨房中，使用简单，且一旦拥有就无法离弃。省时。节能。安全。 在改变日常厨房任务的同时，还能帮您节省时间，减少浪费，并有助于提高安全性。在您自己的家中观看其实际运行情况。预约我们的专家团队进行免费的实时虚拟演示。他们将为您一对一地介绍各项功能并回答任何问题。扫描二维码以解锁专属优惠并安排虚拟演示，或访问 .co.uk。体验中心：Beaumont Buildings, Great Ducie Street, Manchester, M3 1PQ。“ 不仅仅是一个水龙头，它是厨房里的游戏规则改变者，而且这种公司会让你觉得钱花在了正确的地方。”# ® 100℃ 沸水、冰水和气泡水 全能水龙头 安装时间可能因厨房布局而异。

---

Speedmaster

自 1957 年起成为经典 SPEEDMASTER GREY SIDE OF THE MOON 同轴天文台认证 Ω OMEGA

---

[数据参考] 以下原文数字因单位格式转换可能未在译文中完整体现（供参考）：800000、650000、183000、32000、50000、77000

New Scientist

WEEKLY 8 August 2026

AI'S NEXT LEAP

What happens when AI starts to experience the real world? We're about to find out

GENOME OF GHOST ANCESTOR DISCOVERED IN OUR DNA

THE CONTROVERSIAL ALZHEIMER'S SURGERY GIVEN TO THOUSANDS

HOW EUROPE CAN ADAPT TO DAWNING ERA OF WILDFIRES

PLUS

WANDERING BLACK HOLE

UNDERSTANDING GRIEF ATTACKS

GREENER ALTERNATIVE TO AIRCON

No3607 £7.95 CANS11.99

---

MACMILLAN CANCER SUPPORT

"If I had not consistently pushed for answers, I would not be here now." Denzil

For months, Denzil suffered severe abdominal pain with no answers. When tests flagged a low platelet count, his GP suggested it was possibly down to his Jamaican heritage. He kept pushing through multiple visits, until finally a diagnosis - colon cancer.

From day one, Macmillan was there with advice, signposting and support.

But not everyone gets the care they deserve. Whether you do can depend on where you live, what you earn, or simply who you are.

Donate to Macmillan today and you could help make sure our Welfare Rights Advisors, Buddies and Support Line are there for everyone who needs them.

Will you donate £10 and make sure nobody faces cancer without the care they need?

Cancer isn't fair. Cancer care should be. Donate now at macmillan.org.uk/news

Scan this QR code to donate now

Macmillan Cancer Support, registered charity in England and Wales (261017), Scotland (SC039907) and the Isle of Man (604). Also operating in Northern Ireland. MAC20415_NewScientist1

Registered with FUNDRAISING REGULATOR

---

This week's issue

On the cover

30 AI's next leap

Vol 271 No 3607 Cover image: Andrea De Santis

7 Genome of ghost ancestor discovered in our DNA

6 The controversial Alzheimer's surgery given to thousands

4 How Europe can adapt to dawning era of wildfires

9 Wandering black hole 40 Understanding grief attacks 13 Greener alternative to aircon

36 Features "Microbiome profiling could reveal clues about an unknown murderer"

News

8 Amazon's lost civilisation

Aquiry society may have had up to 3 million people

10 Quantum advantages

Quantum computers are starting to become useful as scientific tools

12 Reversing blindness

Eye implant that partly restores vision approved for use

Views

18 Food for thought

Helen Thomson on ultra-processed foods and brains

19 Clumpy cosmos

Is it time to abandon the most central law in cosmology, asks Leah Crane

20 Medicine's new wave

Drugs for "undruggable" proteins with Michael Le Page

22 Aperture

Striking image captures the damage left by fishing nets

24 Book of the week

A new way to think about life

5 Fire clouds Dangerous storms are fanning the flames in France

Features

30 The prediction machine

A new kind of AI is starting to understand cause and effect in the physical world

36 CSI: Microbiome

Can we use hidden microbial clues to solve crimes?

40 Grief attacks

The loss of a loved one can lead to sudden panic attack-like episodes

The back pages

45 Puzzles

Try our crossword, puzzle and quick quiz

46 Almost the last word

How far back do family trees stop doubling and start decreasing?

47 Tom Gauld for New Scientist

A cartoonist's take on the world

48 Feedback

A look into the geometry of balloon animals

48 Twisteddoodles

for New Scientist

Picturing the lighter side of life

8 August 2026 | New Scientist | 1

---

Over

10,000

electric vehicles and counting.

This year, we're increasing the electric vehicles used for deliveries across Europe to over 10,000.

amazon

Based on projected delivery service partner vehicles across the UK & EU by end of 2026.

---

The leader

A burning question

We must tackle the root cause of climate change, not just adapt

WHEN your house is on fire, should you call a firefighter, an architect or a politician? OK, that's an easy one – but what about when the world is on fire? That is a question many of us will be wrestling with as dangerous blazes spring up across much of the world.

Calling a firefighter fixes your immediate problem, which is that your house is burning down. What it doesn't address is why your house is burning, or the steps you could have taken to reduce the risk of disaster.

As we report on page 4, the question of how to prevent wildfires in Europe is now at the top of many minds on the continent. This means changing the way that land is used in order to build in resilience, much as an architect

might design a home with fire doors or other safety features.

Learning to live with wildfires in this way is a form of climate adaptation. It is essential that we redesign our way of life to ease the burden of the warmer world we have created, but that doesn't mean

“As long as our society is dependent on fossil fuels, Earth's climate will continue to warm”

we should ignore the underlying cause: burning fossil fuels and greenhouse gas emissions.

And yet, in responding to Europe's extraordinary summer, some people seem to want to do just that. “We've lost the war against global warming –

so we should ditch net zero and learn to handle hot weather” was the view of one columnist in The Telegraph, a UK newspaper known for publishing climate-sceptic views.

This is entirely wrong, and it is where our politicians must step in. If society is organised in such a way that your house keeps getting set on fire, there is only so much an architect can do – eventually, it will burn down. Similarly, as long as our society is organised around burning fossil fuels, Earth's climate will continue to warm and the wildfires will continue to get worse. Achieving net zero, meaning we no longer emit a meaningful amount of greenhouse gases, is the only way to do that. Politicians who say otherwise might as well hand out matches. ■

PUBLISHING & COMMERCIAL

Display advertising Tel +44 (0)203 615 6456 Email displayads@newscientist.com Sales director Claudia Nicoletti Partnership growth lead Joe Teal Agency partner Daniel Sinclair

New Scientist Events Tel +44 (0)203 615 6554 Email live@newscientist.com Sales director Jacqui McCarron Sales manager Maureen Ignacio Head of product delivery Henry Gomm Events and projects manager Georgia Hill Events and projects executive Clara Lucas

New Scientist Discovery Tours Email tours@newscientist.com Director Kevin Currie Head of travel marketing Ginny Smith Head of travel product Lara Paxton Travel product manager Pip Orchard

Marketing & Data Chief customer officer Claire Overstall Head of campaign marketing James Nicholson Subscription acquisitions manager Molly Hayde Campaign marketing coordinator Charlotte Weeks Head of customer experience Emma Robinson Engagement marketing manager Kelly Spillane Senior marketing executive Sarah Fabian Head of CRM & audience data Rachael Dunderdale Analytics manager Gary Parung Junior data analyst Joseph Cole CRM manager Fion Evans CRM executive Komal Nadeem

Technology Director of engineering Michael Ilett Principal engineers Phil John, Tom McQuillan Senior engineers Amandeep Sian, Maria Garrido, Matt Watson QA engineer Kaveri Daitkar

Product Head of product Remy Becher Product manager Caitlin Jenkins Product manager Thomas Reader Head of product design Irmak Sunal-Izlikin Digital marketing designer Katarina Pollu Brand marketing designer Laura Spanu Senior product designer Sophie Majecki

Partnerships Consultant editor Justin Mullins Editorial director B2B Edward Glanville Head of strategic partnerships David Allard

New Scientist

Chief executive Roland Agambar Chief operating officer Jonas Hermans Chief financial officer Depak Patel Executive assistant Lorraine Lodge

Finance & operations Head of finance Charlotte Lion Management accountant Ilir Morina HR business partner Tinka Bleijenberg

CONTACT US

newscientist.com/contact

General & media enquiries

UK 9 Derry Street, London, W8 5HY Australia 58 Gipps Street, Collingwood, Victoria 3066 US 51 Astor Place, 4th Floor, New York, NY 10003

UK Newsstand

Marketforce UK Ltd Email mfcommunications@futurenet.com

Syndication

Tribune Content Agency Tel +44 (0)20 7588 7588 Email tca-articlesales@tribpub.com

Subscriptions

View the latest offers newscientist.com/subscription Tel +44 (0)330 333 9470 Email subscriptions@newscientist.com Post New Scientist, Rockwood House, Perrymount Road, Haywards Heath, West Sussex RH16 3DH

© 2026 New Scientist Ltd, England. New Scientist is published weekly by New Scientist Ltd. ISSN 0261-4073. New Scientist (Online) ISSN 2059-5387. Registered at the Post Office as a newspaper and printed in England by Warners Midlands Plc

EDITORIAL

Editor Catherine de Lange Executive editor Timothy Revell Managing editor Penny Sarcher

News

News editor Jacob Aron Assistant news editors Alexandra Thompson, Sam Wong Reporters (UK) Madeleine Cuff, Michael Le Page, Alec Luhn, Matthew Sparkes, Carissa Wong (Aus) Alice Klein, James Woodford

Digital

Head of audience Matt Hambly Podcast editor Rowan Hooper Head of editorial video David Stock Senior Social Media Manager Isabel Baldwin Video producer Obomate Briggs

Features

Head of features Claudia Canavan Deputy head of features Joshua Howgego Editors Abigail Beall, Alison George, Olivia Goldhill, Jacklin Kwan, Thomas Lewton

Culture and Community

Head of books and culture Alison Flood Senior culture editor Liz Else

Magazine

Magazine editor Eleanor Parsons Assistant magazine editor Sophie Berdugo

Subeditors

Chief subeditor Kelsey Hayes Bethan Ackerley, Tom Campbell Trainee subeditor Lucy Wall

Design

Art editor Ryan Wills Joe Hetzel, Phoebe Watts

Picture desk

Picture editor Tim Boddy Assistant picture editor Jenny Quiggin

Production

Head of publishing operations Joanne Keogh

New Scientist US

Senior editor Chelsea Whyte Features editor Grace Wade Subeditor Alexis Wnuk Audience editor Gerardo Bandera Reporters Leah Crane, Karmela Padavic-Callaghan

8 August 2026 | New Scientist | 3

---

News

Environment

Europe faces fiery new reality

Paradoxically, suppressing wildfires has also made European countries more vulnerable to huge blazes. Experts say prevention strategies need updating, finds Alec Luhn

AS UNPRECEDENTED wildfires continue to burn in France and Spain, the blame game has also heated up. Right-wing opposition politicians have claimed that both nations' governments under-invested in water bombers because nearly half of their Canadair planes are out of operation. Officials have accused conservative-led regions like Madrid, where the forest fire service is only half-staffed, of undermining fire control.

But experts say the debates over firefighting are focusing on the symptoms, rather than the disease. Instead, Europe must invest more in fire prevention measures, such as intentionally starting controlled fires to manage vegetation. That includes the UK, where a wildfire in Suffolk came within 5 kilometres of the Sizewell nuclear power station and another burned 25 square kilometres of moorland and forest in the Cairngorms in Scotland.

"You can double the number of planes; you can triple it. It will be very, very costly, and every single euro you put there is not a euro that is invested in prevention," says Sebastien Lahaye at the

"Shrubs and brush have proliferated, so those fires that do get out of hand are more likely to blow up"

French wildfire consultancy Warucene. "You will nevertheless have one fire or several fires that will escape and that will become this huge fire."

The ingredients of a catastrophic wildfire are drought, heat and an accumulation of fuel, and this year has mixed in a generous portion of all three. A wet spring in parts of Europe increased vegetation, which then dried out in a widespread drought this summer. Now, much of the continent is suffering

through its fourth heatwave.

More than 4300 square kilometres have gone up in flames across the European Union, a pace that, if it continues, will break the record of 10,800 square kilometres burned last year. Record-sized wildfires near Bordeaux and Madrid forced one of the largest evacuations since the second world war.

In southern Europe, the average number of wildfires each year and the area burned has actually declined since 1981, according to a study published last week by Raúl Cordero at the University of Groningen in the Netherlands and his colleagues. Mediterranean countries have warned the public against campfires and cigarettes in woodlands, and got better at quickly extinguishing blazes by training firefighters and buying planes and helicopters.

But this crackdown has created a tinderbox. Shrubs and brush have proliferated, so those fires that do get out of hand are more likely to blow up into extreme megafires. This "suppression paradox" has been compounded by an exodus of people from rural areas to cities, leaving former farms and pastures overgrown. At the same time, more houses have sprung up in forests outside cities like Barcelona, raising the risk of wildfires, which most often are ignited by humans.

Now, global warming is catching up with the improvements in wildfire suppression. Hot, dry and windy fire-weather days have more than doubled across much of southern Europe, the study found.

"What's occurring this year seemed to confirm our hypothesis... that we are about to enter a new era of large fires that didn't exist in Europe because of the successful fire suppression and education," says Cordero.

CESA MANGOY/FAI GETTY IMAGES

Firefighters battle a blaze in Ávila, which was the largest in Spain's history

Rather than responding with more suppression, Europe must embrace the often-uncomfortable task of making landscapes fire-resilient. Roughly speaking, 90 per cent of the fire budget in many countries goes towards fighting fires and 10 per cent towards preventing them, according to Johann Goldammer at the Max Planck Institute for Chemistry in Germany. "It should be the other way round," he says.

The EU has been working to restore nature in 20 per cent of its land area by 2030 to combat the climate and biodiversity crises. But in many cases, woodlands need to be more intensively managed, including through sustainable logging, Goldammer argues.

Logging alone isn't enough. The fire near Bordeaux raged through Europe's largest human-made forest in Landes, 10,000 square kilometres of land that was once

so swampy that shepherds walked on stilts. Napoleon III reclaimed the marshes by planting maritime pines, which are still harvested for paper and lumber.

Especially since a devastating 2022 fire there, landowners have built access roads, water ditches and fire breaks. But they also replanted the pine monoculture, a practice that raises fire risk. Similar monocultures of eucalyptus in Portugal have fuelled blazes like the Pedrógão Grande fire, which killed 66 people in 2017. While diversifying trees in Landes may be difficult, the fire breaks at least need to be widened and increased, says Lahaye.

Fighting fire with fire

Even where forests don't have fire breaks cleared to the bare ground, trees can be thinned and brush cleared to create fuel breaks. Re-establishing orchards, vineyards, pastures and agriculture can help create

4 | New Scientist | 8 August 2026

---

'Fire clouds' are making wildfires more dangerous

a "mosaic" of higher and lower fire fuel loads, rather than the uninterrupted tree cover that allowed the Bordeaux and Madrid fires to explode out of control.

Spanish regions have deployed herds of goats, sheep and cows to devour flammable brush. But in France, only 30 per cent of risky brush has been cleared due to conflicts between wildfire regulations and biodiversity regulations.

The most effective way to clear undergrowth is often through prescribed burns. Wildfires in the western US in 2020 were 16 per cent less severe in areas that had previously undergone prescribed burns, and they emitted 14 per cent less harmful particulate pollution as well.

But Europe is much more densely populated, and locals often see the technique as a threat. In some areas, regulations still often prevent it. While prescribed burning has been conducted along roads in Landes, it is excluded from the timber plantations. ■

A PYROCUMULONIMBUS cloud, a thunderstorm generated by an intensely hot wildfire, has been observed with radar for the first time in France. These deadly "fire clouds" of rising hot air, smoke and moisture, which accelerate blazes with strong winds and soaring embers, are expected to become more common in Europe.

"You have this tilted column that is sucking air and then propagating embers in front, and it also creates lightning. That's another way of lighting up fuels in front of the fire, so these cells are unstoppable," says Jean-Baptiste Filippi at the University of Corsica, France.

The wildfire burning in the Gironde region of France, close to the city of Bordeaux (see main story, left), generated a pyrocumulonimbus cloud on 24 July and another on 25 July, according to regional officials. Embers and lightning ignited new fires around the original inferno.

Sometimes called the "fire-breathing dragon of clouds",

A fire cloud seen near the village of Saumos in France

pyrocumulonimbus are typically generated by massive wildfires in the forests of North America, Siberia and Australia. They have spurred on infamous fires like the Black Saturday bushfires that killed 173 people in Australia in 2009.

In Europe, however, there have been only a handful of confirmed pyrocumulonimbus in Spain and Portugal, including

"Dozens of lightning strikes can ignite blazes as far as 100 kilometres ahead of the main fire"

one during Portugal's devastating Pedrógão Grande fire in 2017, which destroyed hundreds of homes and killed 66 people. But that appears to be changing.

"You have to have two things happening at the same time: an extreme heat source and an unstable atmosphere that interacts with the fire," says Francesca Di Giuseppe at the European Centre for Medium-Range Weather Forecasts. "If... the heat from the fire increases, like it's projected to increase from our studies, then yes, you should see more of these events."

This firestorm begins to take shape when temperatures as high as 800°C send a plume of smoke into the sky. As the hot air rises and cools, water vapour condenses onto the ash to form pyrocumulus clouds.

If the plume is hot enough and the atmosphere is unstable enough, the cloud can keep rising to reach the upper troposphere at 10,000 metres, the cruising altitude of a passenger plane. It sometimes reaches the stratosphere, spreading as much smoke as a moderate volcano. It becomes a pyrocumulonimbus when the static electricity from colliding ice particles sparks thunder and lightning.

Although rain from the cloud often evaporates in the hot air, this kind of thunderstorm nonetheless creates its own weather system. The cloud's updraft pulls in air, fanning the flames with winds reaching 160 kilometres per hour and accelerating the fire's spread by up to five times, according to Filippi.

Meanwhile, strong downdrafts create erratic gusts that carry embers off to start new fires. Dozens of lightning strikes can ignite blazes as far as 100 kilometres ahead of the main fire. It can also create fire tornadoes of twisting flames. As the winds, embers and lightning strengthen and expand the wildfire, it releases more heat and smoke, a feedback loop that can amplify the rising cloud for many hours.

Both wind shear higher in the atmosphere and erratic gusts at the ground can take the fire in unpredictable directions, making it nearly impossible to contain its acceleration or plan targeted evacuations. ■ AL

© 2017 SHANDOLPH/PA/SHUTTERSTOCK

8 August 2026 | New Scientist | 5

---

News

Neurology

Questions over dementia surgery

Thousands of people in China have had a now-banned experimental procedure that aims to help the brain flush away Alzheimer's-linked proteins, reports Carissa Wong

MORE than 2000 people in China are thought to have undergone an experimental surgery to treat Alzheimer's disease, which allegedly helped a man with the condition who was bedbound walk again. The procedure – which involves connecting the lymphatic system to nearby veins to remove Alzheimer's-associated proteins – has now been banned in the country.

"When I first heard of this treatment, I was shocked," says Steven Proulx at the University of Bern in Switzerland. "There is basically no evidence supporting this."

But initial results from small safety trials in the US and Singapore, which were presented earlier this month at the Alzheimer's Association International Conference in London, suggest the approach is safe for up to six months, with some signals of cognitive improvement.

An ongoing randomised trial is now comparing the surgery with a sham procedure, which should provide a clearer picture of its potential, but some have questioned the ethics of this and how helpful the trial will be.

The surgery, called deep cervical lymphovenous anastomosis (DCLVA), aims to enhance the drainage of brain waste via the lymphatic system, a network of vessels and nodes that collects waste from tissues to be broken down or removed from the body. The idea is that surgically attaching this network to nearby veins accelerates this process to more efficiently clear Alzheimer's-related proteins like beta-amyloid and tau.

A computed tomography (CT) scan showing the brain of a woman with Alzheimer's

It was developed by Xie Qingping at Hangzhou Qiushi Hospital in China, based on a routine treatment for chronically swollen limbs. In 2024, Qingping – who didn't respond to New Scientist's requests for comment before going to press – and his colleagues reported in the journal Plastic and Reconstructive

"It is unclear whether any improvements were substantial enough for people to notice them"

Surgery that they had conducted DCLVA on 50 people with Alzheimer's disease. They also released a video of a man with Alzheimer's who had been bedbound briskly walking eight months after the surgery.

But some have concerns over the reliability of this case report. "A video from one patient I would not find trustworthy," says Proulx,

who added that the researchers didn't assess for other improvements, such as changes to the man's brain-imaging scans.

Still, the footage went viral in China, says Fan Fan at Augusta University in Georgia. By May 2025, thousands of people were estimated to have had the procedure in around 300 private and public hospitals across China.

Reports suggest that many of these were done by plastic surgeons, says Proulx, with fees ranging from tens of thousands to several hundred thousand Chinese yuan (thousands to tens of thousands of pounds).

Lack of evidence

But in June 2025, China's National Health Commission banned the procedure outside of research settings, citing a lack of evidence for safety and effectiveness. The

commission didn't respond to New Scientist's request for comment before going to press. "It was appropriate for China to stop DCLVA from being offered for Alzheimer's disease," says Shiju Gu, who previously researched Alzheimer's disease at Harvard University.

To date, about a dozen case reports and trials have reported doing DCLVA on people with moderate or severe Alzheimer's. The US and Singaporean trials presented at the recent Alzheimer's conference suggest it is safe over the short term. The Singaporean trial also found evidence of small cognitive improvements in two out of its four participants.

These improvements aren't necessarily trivial, says Jeff Iliff at the University of Washington in Seattle, but none of these trials was randomised or blinded.

"Based on the evidence currently available, I would describe DCLVA's effectiveness as uncertain because the existing clinical studies in China have generally been uncontrolled, short term and methodologically limited, making it difficult to separate a real surgical effect from placebo effects, postoperative care or fluctuations in the patient's condition," says Gu.

It is also unclear whether any improvements were substantial enough for people or their families to notice them, says Per Kristian Eide at the University of Oslo in Norway.

A randomised controlled trial comparing DCLVA with sham surgery is now ongoing. Led by Hui-Sheng Chen at the General Hospital of Shenyang Military Region in China, who didn't respond to New Scientist's request for comment before going to press, the trial is aiming

6 | New Scientist | 8 August 2026

---

Human evolution

We all have DNA from 'ghost' ancestors in our genomes

Michael Le Page

to recruit 50 people with mild to moderate Alzheimer's. Results are expected next year.

Ethical concerns

Despite previous trials being criticised for not having control groups, some have questioned the ethics of sham surgery. "The risks of any kind of surgery and the accompanying anaesthesia in an elderly and potentially frail population makes the performance of a 'fake' intervention very difficult to justify, despite the importance to the overall study," says Iliff.

Eide, who shares these ethical concerns, also points out that the trial's protocol says it will look for cognitive improvements four days after the surgery. "You wouldn't expect any improvement in such a short time," he says.

Yet the surgery could have one major advantage: it is a one-off intervention. That makes DCLVA worth pursuing, says Vincent Tay at Changi General Hospital, who is involved in the Singapore trial.

Treatment for Alzheimer's generally focuses on relieving symptoms, but antibody therapies such as lecanemab that remove beta-amyloid clumps from the brain are approved in the US and the UK. But their questionable efficacy and high cost, partly due to being infused, mean they aren't available through the UK's health services. "The upside is this [surgery] is a one-time thing," says Raag Airan at Stanford University in California.

So, could this procedure one day be routine in China, and even around the world? "I'm paying close attention to results being published in this space, as is everyone working on this question," says Iliff. ■

© ENTERPRISE/SEARCH/SEARCH PHOTO LIBRARY

IN ANY person alive today, between 0.5 and 1 per cent of their DNA comes from an unknown hominin species that interbred with Homo sapiens before they left Africa.

Different people have different bits of DNA from this "ghost" ancestor and, crucially, they are spread fairly evenly around the genome. This means that, if all the bits could be put together, it might be possible to assemble almost an entire genome of this ancient species.

Earlier research using ancient DNA sequences from fossils revealed that Homo sapiens interbred with Neanderthals and Denisovans within the past 50,000 years.

While previous work has also identified ghost ancestry from at least two other lineages, Priya Moorjani at the University of California, Berkeley, and her colleagues used a new method that is much more powerful. It is based on constructing a "family tree" for parts of the genome, which has only recently become computationally possible.

"I believe it is the next frontier

of research in human genetics, because it allows us to even learn about our ancestors for whom we don't have any fossil [DNA] evidence," says Moorjani.

Only a small part of the Neanderthal and Denisovan genomes remain collectively preserved within us. "Long segments of the genome are devoid of, or have very low

"It might be possible to assemble almost an entire genome of this ghost ancestor"

fractions of, Neanderthal or Denisovan ancestry," says team member Arjun Biddanda at Johns Hopkins University in Maryland. "Yet we were able to find high-ish fractions of this ghost ancestry in these regions."

The researchers think this is because the populations of Neanderthals and Denisovans were very small, and therefore were more likely to have detrimental mutations. "Most likely there was selection against Neanderthal or Denisovan ancestry in

Homo heidelbergensis may be the source of a small fraction of our DNA

these regions," says Moorjani.

But much more of the DNA deriving from one ghost ancestor appears to have been selected for, or it was at least neutral in evolutionary terms. This is probably because the population of this ghost ancestor was much larger, she says.

The team's work suggests that the ghost ancestor split away from the main lineages leading to modern humans around 800,000 years ago. Then, some time later, the two lineages came back together and interbred (Science, doi.org/rhtb).

"Because all modern humans carry signatures of the ghost ancestry, it's most likely that the gene flow occurred in Africa, and that this population existed in Africa," says Moorjani. Based on all these factors, this ghost ancestor could be a species known from fossils, such as Homo heidelbergensis, the researchers speculate.

A second ghost ancestor identified by their approach is much older. It split from the main lineage leading to modern humans around 1.8 million years ago, but later interbred with Denisovans in Eurasia. Modern humans inherited its DNA via Denisovans, says Moorjani.

The most likely candidate for the older ghost lineage has to be Homo erectus, says Chris Stringer at the Natural History Museum in London.

"These results are exciting," says Stringer. "They provide indirect ways of reconstructing parts of the genomes of more primitive species like Homo erectus and Homo heidelbergensis." ■

8 August 2026 | New Scientist | 7

---

News

Archaeology

Amazon's huge lost civilisation

The little-known Aquiry society built tens of thousands of giant earthworks in the rainforest and had a population of up to 3 million people, finds James Woodford

SURVEYS over the dense Amazon rainforest using canopy-penetrating lasers have revealed hundreds of vast structures built by a lost civilisation that thrived for almost 1500 years. The findings lend weight to suggestions that millions of people once lived in areas today regarded as wilderness.

The Aquiry civilisation, which lasted from roughly 600 BC to AD 850, built vast geometric earthworks, whose shapes may represent various deities, without using metal tools or animals. Many are also connected by ancient roads.

These people were little known to science until recent decades because the part of Amazonia where they lived constituted what was thought to be "an almost uninhabited pristine rainforest", says Martti Pärssinen at the University of Helsinki, Finland. But researchers have been learning more about the civilisation's extent using lidar, a laser-scanning technique that can create a three-dimensional map of the ground under most vegetation.

From aircraft flying over the forest, around 30 laser pulses are shot at the ground every square metre. Small gaps in the forest canopy allow some of these pulses to reach the terrain below and return to the aircraft.

In 2024, Pärssinen and his colleagues flew over an area of nearly 4500 square kilometres, in the Brazilian states of Acre and Amazonas. The survey revealed 432 ditched geometric enclosures called geoglyphs, nearly 400 of which were completely new to science.

By extrapolating these results across the 183,000 square kilometres of territory inhabited by the Aquiry civilisation, the

The Aquiry geoglyphs discovered in Brazil (left) were revealed using canopy-penetrating laser scans (above)

team estimates there may be as many as 30,000 geoglyphs.

On average, the earthworks cover 3 hectares, but they can be as large as 50 hectares. The ditches are typically about 12 metres wide and up to 5 metres deep. "The scale and precision of these constructions is simply astonishing," says Pärssinen.

Each of the average-sized geoglyphs would have required around 300 people to build and maintain, the researchers estimate. The findings suggest a total population of between 1.25 million and 3 million people at the civilisation's peak between AD 100 and 300, says Pärssinen, and his feeling is that the upper end of the estimate is most likely.

"The notion of Amazonia as an untouched wilderness is largely a myth and [the truth is] that ancient human societies played an important role in

"Each of the average-sized geoglyphs would have required about 300 people to build and maintain"

shaping the present-day tropical rainforest," he says.

Over the past 20 years, Pärssinen and his colleagues have excavated more than 40 ceremonial and settlement sites in Acre, Amazonas and Rondônia (another Brazilian state), as well as in Bolivia.

Mind-blowing magnitude

This showed that the geoglyphs were ceremonial centres used as "venues for gatherings, political decision-making and displays of chiefly generosity, ritual dancing, music, feasting, athletic competitions and ball games", he says. Long dwellings that housed several families were probably situated in a dispersed pattern around the earthworks (Nature, DOI: 10.1038/s41561-026-02054-6).

The Aquiry had a major impact on the Amazon rainforest, says Pärssinen, burning bamboo forest to make way for maize, squash and manioc, as well as clearing forests for the huge earthworks and the extensive

roads leading to them. The Aquiry also grew many fruit-bearing tree species, including the Brazil nut tree and peach palm. It isn't known what caused their demise.

Michael Heckenberger at the University of Florida says the claims about population size are bold, but they are "plausible and extremely important, even mind-blowing".

"The idea of pre-Columbian urbanism in southern Amazonia has gained momentum over the past two decades. What we now need to know is what is going on around the detectable earthworks, such as occupational sites to support the high estimated population," he says.

Charles Clement at the National Institute of Amazonian Research in Manaus, Brazil, says the new research "puts a large-scale culture in the Amazon forest".

"Needless to say, they have not ground-truthed the entire set of new geoglyphs, but previous lidar studies have been remarkably efficient, so I feel that their numbers are solid," he says. ■

8 | New Scientist | 8 August 2026

---

Astronomy

Extraordinary supermassive black hole caught straying

Leah Crane

THE most extreme example of a so-called wandering supermassive black hole has just been spotted.

Black holes don't emit light, so Robert Stein at the University of Maryland and his team found this one only when it devoured a star in what is known as a tidal disruption event, or TDE. A machine learning algorithm identified the burst of light as a possible TDE in data from the Zwicky Transient Facility in California, and the researchers performed follow-up observations with several other telescopes to confirm it.

This TDE is the most distant from the centre of its host galaxy that we've ever seen. It is more than 30,000 light years from the centre of its galaxy – for comparison, our solar system is about 26,000 light years from the centre of the Milky Way (The Astrophysical Journal Letters, doi.org/hccvkm).

Wandering supermassive black holes usually begin their journeys away from a galactic centre due to a merger: two galaxies slam together and one of their central black holes is flung away. Finding these black holes can tell us about the history and growth of their galaxies. "What's exciting about this result is that we think it's the tip of the iceberg," says Stein. "If we can understand how many wandering supermassive black holes there are, then it can tell us about how many mergers the galaxy has been through and how it's evolved over time."

There is a limit to how much we can learn about this particular black hole and its environs. The light from the star being devoured is still too bright to see much of anything else around it.

There is probably a small cluster of stars that is gravitationally tied to the black hole, and one of those caused the TDE as it fell in, but we can't confirm that until the light fades. ■

Health

Could eating less protein extend our lifespans?

Carissa Wong

PROTEIN-RICH products – from cereal to crisps to coffee – are all over supermarket shelves, but mounting evidence in non-human animals suggests that restricting our intake could slow ageing and extend lifespan.

According to the latest review of the evidence, "increasing dietary protein, beyond what's needed, for sedentary individuals or animals leads to metabolic problems and can lead to shorter lifespan", says Dudley Lamming at the University of Wisconsin-Madison.

Lamming and his colleague Bailey Knopf, also at the University of Wisconsin-Madison, sifted through more than 300 studies, mostly in sedentary mice and rats, that explored how restricting protein intake to the minimum required to prevent deficiency affects ageing and lifespan.

In these studies, rodents generally either ate the equivalent of the minimum protein requirement recommended by US guidelines, 0.8 grams of protein per

Upping protein intake may only be beneficial for active individuals

kilogram of body weight, or about double that. The latter is similar to the intake recently recommended by US dietary guidelines for optimal health, 1.2 to 1.6 grams of protein per kilogram of body weight.

In one study, rats that ate around the minimum protein intake lived for nearly 120 days, or more than 50 per cent, longer than those that ate about double that (Cell Press Blue, doi.org/rh2n).

This may be because protein restriction increases levels of the hormone FGF21, says Lamming. In rodents, this acts on the brain, fat tissue and the liver to improve blood sugar levels and reduce the number of senescent cells. These cells, which accumulate with age, have stopped dividing and cause chronic inflammation.

Restricting protein intake has also been shown to dampen the activity of a protein called mTORC1 in mice. mTORC1 ordinarily stimulates cell growth, but reducing its activity boosts cellular repair. "It essentially switches cells from a proliferative state, where they're growing and engaging a lot of metabolism, to one where

they're dividing more slowly, if at all, and focusing on repair," says Lamming. Studies suggest this cuts the risk of age-related conditions and extends lifespan in mice, he says.

All this suggests that the recent trend to up our protein intake is probably fine among people who exercise regularly, because they may need more protein to repair muscle

"Increasing dietary protein, beyond what's needed, for sedentary individuals leads to metabolic problems"

damage, but eating more than is needed could cause harm in sedentary individuals, says Lamming. But Donald Layman at the University of Illinois Urbana-Champaign points out that rodents vary hugely from people when it comes to protein metabolism. Protein restriction may therefore affect mTORC1 pathways differently between the different species, he says.

What's more, the review doesn't include most of the studies in people that suggest upping our protein intake beyond the minimum cuts our risk of frailty, he says.

Lamming says no studies were intentionally omitted. Beyond building muscle and preventing frailty, there is mixed evidence on whether a higher protein intake helps with ageing, he says.

For instance, one study linked this to a reduced risk of age-related muscle loss, or sarcopenia, among older twins in the UK. But the rates of sarcopenia were so low that the link is unreliable, says Elizabeth Williams at the University of Sheffield in the UK. ■

CARLOS GERRON/SEYNE-HAGUE

8 August 2026 | New Scientist | 9

---

News

Technology

Quantum advantages multiply

Three problems that conventional computers couldn't crack have been solved by quantum computers, showing their potential importance for science, says Karmela Padavic-Callaghan

QUANTUM computers are pushing further into the world of the impossible. Three new examples of calculations that no other type of machine can do show how such computers are becoming useful as scientific tools.

There are many situations in science where solving a problem requires extensive computer calculations, such as when trying to predict the behaviour of large molecules. But in some cases, the calculations are so complex that traditional computers will never reach an answer.

This is the point at which quantum computers can have "quantum advantage", as they can use quantum phenomena to produce a solution. But how do you know whether you can trust an answer that only one machine can produce?

Researchers at IBM and their colleagues have now claimed three new cases of quantum advantage, together with arguments for why these calculations should be trusted.

Two of the cases were simulations of mathematical models that are of particular interest to physicists who study materials. In one, the model aims to capture how quantum materials behave when they are hit by laser light, focusing on a type of magnet. Using a quantum computer, the researchers discovered that some of the magnet's properties oscillate, which wasn't known before. Notably, simulations on the Fugaku supercomputer in Japan, as well as on an Nvidia GPU, failed to produce results.

In the second case, the team focused on how information

IBM researchers have used simulations to test quantum machines

gets scrambled within a very uneven material, which could help researchers to better understand the chaotic processes that take place when, for example, a catalyst is used to accelerate a chemical reaction.

Here, researchers compared the quantum computer's output with a leading simulation method for traditional computers – and one that has dispelled quantum advantage claims in the past.

"You can actually take the data from the experiment and get an estimate for how good the runs were"

For some simulation parameters, these methods couldn't measure up to the quantum computer.

For both simulations, the team made an effort to estimate how much the quantum computer could be trusted by repeating the simulation on several different quantum computers, each of which had slightly different

amounts of noise, and they also deliberately injected noise into some computational runs. The comparison between these differently noisy computations enabled the team to quantify how much of the quantum computer's output can be trusted and how much is just noise, says Abhinav Kandala at IBM.

The third example focused on a sampling problem that looks at the output of many quantum-computing circuits that have pre-specified properties, such as the distribution of mathematical operations within them. Researchers already knew that sampling problems are difficult for conventional computers, but even though a quantum computer could output an answer, it previously struggled to verify how meaningful it was – whether it was mired by errors. The team devised a procedure that also involved tweaking and then repeating the computation to quantify the fidelity of the

quantum computer's answer.

Quantum computers are now operating like we have always wanted them to, says Kandala. "I hope people begin to use them as scientific instruments that have trusted outcomes."

Creating a toolbox

These new efforts to verify just how well quantum computers are doing are a step in the right direction, says Dominik Hangleiter at ETH Zurich in Switzerland, who wasn't part of the team. "You can actually take the data from the experiment and get an estimate for how good the experimental runs were." This process is still not ideal, as it isn't fully agnostic of the quantum computer's specifics, but it is necessary for the still fairly small and noisy quantum computers we have today, he says.

In addition to further improvements on the verification front, quantum computer simulations of materials also need to be pushed further. Emanuele Dalla Torre at Bar-Illan University in Israel says the magnet simulation made findings significant from a fundamental-science perspective, yet it was still idealised compared with real-world magnets. "The problem itself is, at the end, a basic science problem, which doesn't [necessarily have] a clear usage right now."

But the team is optimistic about how these new developments could jump-start an era of quantum usefulness. "We are creating the toolbox for how we can trust the output [of quantum computers]. And once we can trust the output, then we can start to nail down applications," says Jay Gambetta at IBM. ■

COWNE JINGUIN

10 | New Scientist | 8 August 2026

---

Archaeology

Smallpox in 400-year-old Chilean mummies came from Europe

James Woodford

DNA samples collected from Inca mummies have confirmed the presence of a strain of smallpox that first emerged in Europe.

"We interpreted this as the first direct molecular evidence of smallpox introduction into the Americas through European colonisation," says Shigeki Nakagome at Trinity College Dublin, Ireland.

Smallpox was the scourge of humanity for centuries until its eradication in the 1980s, through vaccination, from all but a few high-security medical laboratories. Severe strains of the virus had a mortality rate of up to a third and left survivors with scars from the blisters and scabs that cover the skin and fill with pus.

The disease was particularly brutal in parts of the world colonised by Europeans, such as Australia and the Americas, where Indigenous populations had no natural immunity, resulting in millions of deaths.

In the Americas, historical reports record that smallpox began devastating Indigenous

communities in the early 16th century, after the arrival of European colonists.

Matthew Ward at the University of Dundee in the UK, who wasn't involved in the new research, says the waves of epidemics, including smallpox, that spread across the Americas were some of the most important events in history. "These diseases killed perhaps 95 per cent of

Francisco Pizarro led the Spanish conquest of the Inca Empire

FOTOCA B. ARSKETT MAGS

the Indigenous population of the Americas and allowed for European colonisation," he says.

Nakagome and his colleagues studied samples of bone from mummies found at an archaeological site known as Camarones 9 in northern Chile, discovered in 1957. Over subsequent decades, 41 mummies wrapped in blankets and buried with grave goods were recovered from the site.

"While the original aim was to study the human DNA, we also searched for DNA from ancient pathogens," says Nakagome. "That is when we discovered fragments matching the smallpox virus in two individuals."

The first was a woman who was 30 to 35 years old at the time of her death, and the second, an 18-to-20-year-old man, was revealed by carbon dating to have died between 1492 and 1631.

By reconstructing the virus's genomes and comparing them with other ancient and modern smallpox viruses, the team found that they belonged to a lineage of smallpox that

includes early medieval and early modern European strains (Science, doi.org/rhsr).

The discovery adds weight to the possibility that small skin lesions on the trunks of the mummies, previously attributed to chronic arsenic exposure, may in fact have been the distinctive pustules that scar the skin of smallpox victims.

"This finding allowed us to identify a small group of people impacted by the virus for which no historical record exists," says team member Constanza de la Fuente Castro at the University of Chile in Santiago. Before the new research, the pathogen was known only from historical accounts of large-scale outbreaks, she says.

The work was particularly sensitive because of the disturbance of the original remains and the loss of material during analysis, both considered extremely controversial subjects to Indigenous people and the scientific community. "We never forget that behind every DNA extraction is a person, not a data point," says de la Fuente Castro. ■

Health

Guillain-Barré drug switches off part of immune system

THE first targeted treatment for Guillain-Barré syndrome has worked well in a late-stage clinical trial.

"Patients got better sooner and more completely on every measure in our studies," says Doug Love at Annexon Biosciences, the Californian company developing the treatment.

Guillain-Barré syndrome is a rare condition that can affect anyone who has recently had an infection. After fighting the infection, a person's

immune system starts mistakenly attacking their nerves. This assault causes weakness or tingling in their arms and legs, which can progress to difficulty walking, speaking, swallowing and breathing, and sometimes paralysis or death.

The new treatment, called tanruprubart, is a monoclonal antibody that works by blocking a protein in the blood called C1q. This switches off a branch of the immune system called the classical complement pathway that is misdirected to attack nerves in Guillain-Barré syndrome.

In a late-stage trial in Bangladesh

and the Philippines, 241 people with a recent diagnosis of the condition were given an intravenous infusion of tanruprubart or a placebo infusion. On average, those treated with the drug required 28 fewer days of mechanical ventilation and seven fewer days in intensive care, and were able to walk independently 31 days earlier.

The participants who received tanruprubart experienced greater

"On average, those treated with the drug required 28 fewer days of mechanical ventilation"

quality of life and mobility after two months compared with those who had the placebo. The results were presented at the Congress of the European Academy of Neurology in Geneva, Switzerland, in June.

No serious side effects occurred in the trial, and even though tanruprubart switched off part of the immune system, it didn't appear to make the participants more vulnerable to infection.

Based on these results, Annexon Biosciences recently submitted tanruprubart for approval from the European Medicines Agency. ■ Alice Klein

8 August 2026 | New Scientist | 11

---

News

Health

Eye implant partly restores sight

A device that lets people with severe vision loss read has been granted EU and UK approval

Oluoma Aneke

A RETINAL implant that improves vision in people with a common form of blindness has been approved for use in Europe. This is the first time that such a device – which enables people with a type of severe vision loss to read – has been approved anywhere in the world.

“We’ve had patients who could recognise the face of the queen and others who’ve drawn the Sydney Opera House through the device,” says Darius Shahida at Science Corporation in Alameda, California, the company behind the implant.

Called the PRIMA implant, the device restores functional central vision – the straight-ahead sight required to see fine details and read – in people with geographic atrophy. This is a progressive and irreversible form of dry age-related macular degeneration, which involves the slow deterioration of a part of the retina called the macula as retinal cells die off and aren’t renewed.

Existing treatments for geographic atrophy are extremely limited. In the US, two medicines – given as shots in the eye – are approved that block proteins that may damage the retina.

The implant – which is about

The implant is about the size of a pinhead and is inserted beneath the retina

the size of a pinhead and is inserted beneath the retina – is used alongside glasses with a frame-mounted camera. This captures and processes images, then projects the result onto the implant’s chips using near-infrared light. We can’t see in this wavelength, so the process doesn’t interfere with any existing vision. These chips then convert the

near-infrared light into electric pulses to stimulate neurons called bipolar cells, which are typically not damaged by macular degeneration, in the retina. Ordinarily, bipolar cells work with rod and cone cells, which convert light into electrical signals and are affected by macular degeneration. The PRIMA system bypasses rod and cone cells so bipolar cells can create electrical signals that can be interpreted by the brain.

The glasses generate electricity directly from light, similar to solar panels. “This means patients don’t need a tethered power source,” says Shahida.

In a small study published in October last year, PRIMA restored central vision and improved visual acuity in about 80 per cent of cases. “Normal vision in the US is 20/20 and blindness is 20/200,” says Frank Brodie, also at Science. “Patients using PRIMA reached about 20/160, so they went from blind to not blind.”

Most of the side effects were related to the surgery and resolved within two months. “The physical recovery is quick, and some people feel better within a day,” says Brodie.

Before the intervention is widely used, more research is needed on a larger group of people with long-term safety data, says Sunir Garg at Wills Eye Hospital in Philadelphia, Pennsylvania. The researchers also need to uncover if the implant only benefits people with certain patient- or condition-related characteristics, he says.

The intervention has received a CE certification mark under the European Union Medical Device Regulation, which enables it to be used in the European Union and the UK. PRIMA has also received Breakthrough Device and Humanitarian Use Device designations from the FDA, “which is an important pathway toward approval in the United States”, says Shahida. ■

Palaeontology

An ichthyosaur had a very unexpected last meal

A FOSSIL found in Australia’s outback is the first-ever evidence that ichthyosaurs, large marine reptiles, consumed pterosaurs, the earliest known vertebrates to evolve powered flight.

Recreational fossil hunters discovered some of the ichthyosaur’s fossilised bones in 2019 at a designated public site near the town of Richmond, in a semi-arid area of Queensland’s outback. During the Early Cretaceous, between 120 million and 95 million years ago, the area was part of the Eromanga Sea, a shallow inland

sea that covered over one-third of the Australian continent.

When palaeontologists excavated the site three years later, they uncovered a huge, partially articulated ichthyosaur. Identified as Platypterygius australis, it was almost 7 metres long, the size of an adult orca.

“What startled me was not only the size of the animal, but the way its skeleton had been damaged,” says Matt White at the University of New England, Australia, who led the investigation. He explains that the vertebrae were split, ribs were missing and numerous bones

The ichthyosaur ate the pterosaur before being gobbled up itself

displayed large crushing bite marks.

The only explanation, the team concluded, was that a powerful predator had dismembered and chewed the ichthyosaur, most likely the giant pliosaur Kronosaurus queenslandicus, says White.

The team also found globular masses near the ichthyosaur’s skull. Mechanical removal of rock around the fossil and 30 imaging revealed that one of these globules contained fragments of fish, squid-like cephalopods, and, remarkably, two pterosaur bones, including a jaw fragment. The researchers think this was the ichthyosaur’s gut contents, which probably fossilised near the skull due to its digestive tract spilling out (Gondwana Research, doi.org/hccx5).

“It is almost certainly a case of predation or scavenging, and the first-ever example of a pterosaur consumed by an ichthyosaur,” says David Unwin at the University of Leicester, UK. ■ James Urquhart

12 | New Scientist | 8 August 2026

---

Analysis Environment

Is district cooling the solution to extreme heat in cities? Pipes carrying cooled water are more efficient than air-conditioning. Could they help to chill heatwave-stricken cities in the US and Europe, asks Alec Luhn

MORE than 77 million people in Europe enjoy district heating, where hot water is piped from a central plant to homes and buildings. The US has hundreds of these networks, mostly in multi-building campuses like airports, universities and business districts.

But as both Europe and North America swelter through heatwave after heatwave, many are wondering whether the mirror image, pipe networks of chilled water, could be a better cooling option than air conditioning.

"District cooling systems operate at twice the efficiency, or half the operating expense, of a typical in-building chiller plant," says Rob Thornton at the International District Energy Association. "There's emissions reductions, lower costs of manpower and... by concentrating the function to a central plant, you can serve dozens if not hundreds of buildings."

Heatwaves are getting stronger, longer and more severe around the world, driving a boom in air conditioner sales. But in Europe, the fastest-warming continent, only about 1 in 5 households have air conditioning, and the lack of cool spaces can be deadly in an ageing population.

Air conditioning, while necessary for hospitals, schools and care homes, worsens planet-warming carbon emissions and urban heat islands. It isn't a long-term solution, according to a 2025 report by the European Union Covenant of Mayors, an alliance of local governments. District cooling, on the other hand, can both cool cities and cut emissions, the group said.

Middle Eastern and Asian cities that face scorching heat for months of the year have already begun to build gigantic district cooling networks. Dubai's state-

Some district cooling networks already exist in places like Dubai

backed district cooling company now pumps around 6 gigawatts of refrigeration – equivalent to freezing 1.5 million tonnes of ice per day – to more than 1700 buildings.

The US has about 500 of these district cooling networks, mainly on university and hospital campuses, according to Thornton, and Europe has about 200, mostly in cities in France and Sweden.

These cooling systems typically cool water to 6°C or 7°C in refrigeration tanks and pipe it to heat exchangers in each building in the network. These work like radiators in reverse, cooling the air in a home through contact with the cold water.

A simulated district cooling network of 37 public and commercial buildings consumed 15 per cent less energy and lowered operational costs by 10 per cent compared with individual air conditioning, according to a study in Hong Kong. The Paris district cooling system has cut electricity

and emissions by half compared with individual cooling.

Office buildings often install air conditioning systems that are two to three times larger than is typically needed so that the AC doesn't go out on the hottest day of the year, says Thornton. District cooling can shrink that huge pool of excess capacity.

To increase efficiency even further, district cooling systems in Geneva and Paris chill their liquid

"District cooling systems operate at twice the efficiency, or half the operating expense"

with water from Lake Geneva and the Seine, respectively, while one in Heerlen, the Netherlands, draws on flooded coal mines. In the US, some district cooling networks make ice at night, when electricity is cheap, and pipe it out once it melts during the day.

While most district cooling networks still serve public or commercial buildings, a few, like the one in Geneva, are starting to expand to residential buildings. But district cooling can cost

hundreds of millions of dollars to build, mainly because streets and pavements need to be dug up to lay pipes. It is cheapest if it can be installed alongside an existing district heating system, according to Wangda Zuo at Pennsylvania State University.

"It works for places with really dense buildings that have high demands, and ideally if you already have the tunnels to lay out the piping," he says.

Two-in-one

But a newer technology called an ambient loop system, which combines district heating and cooling in one, could enjoy more uptake in Europe. Upscale projects that include both commercial and residential buildings have installed this technology, but for wider adoption, government backing will probably be needed, says Zuo.

The approach is a way to supercharge individual electric heat pumps, which typically cool or heat air they take from outside the home. An ambient loop system circulates water at whatever temperature the pipes are currently at, usually 10°C to 25°C (50°F to 77°F), although it must occasionally heat it in winter.

Then water-based heat pumps in each building cool or heat this liquid. So, rather than cooling outside air from 40°C to 20°C (104°F to 68°F) during a heatwave, for instance, the heat pump can convert water from 25°C to 20°C, which requires far less electricity.

"If you want to carry something from the foothill of a mountain to the top of a mountain, you have to work very hard," says Zuo. "If the difference in height is low, it's easier. With the ambient loop, the temperature difference is relatively smaller." ■

8 August 2026 | New Scientist | 13

---

Here for every stage

We could help you feel ready for today, and whatever comes next.

Savings & Mortgages

ybs.co.uk

10.6k+ reviews

Trustpilot

Visit us in branch

YORKSHIRE BUILDING SOCIETY

Since 1864

Yorkshire Building Society is a member of the Building Societies Association and is authorised by the Prudential Regulation Authority and regulated by the Financial Conduct Authority and the Prudential Regulation Authority. Yorkshire Building Society is entered in the Financial Services Register and its registration number is 106085. Head Office: Yorkshire House, Yorkshire Drive, Bradford BD5 8LJ. ybs.co.uk

YBM 13893 17 07 26

---

News

Health

'Hypoxia in a pill' may ease symptoms of neurodegenerative conditions

Ute Eberle

A PILL that simulates the effects of high elevation and low oxygen on the brain seems to have eased neurological decline in mice with three neurodegenerative conditions, including Parkinson's disease.

"With 'hypoxia in a pill', we can now manipulate the very air we breathe," says Vamsi Mootha at Massachusetts General Hospital.

While severe altitude sickness can be deadly, anecdotal evidence suggests that symptoms of neurological conditions like Parkinson's disease ease at high elevation, where the air is thinner.

For the past decade, Mootha and his colleagues have done studies in which mice with models of conditions like Parkinson's are kept in air with 11 per cent oxygen, matching the environment at roughly 5000 metres above sea level. They have found that

this can halt, prevent or even reverse symptoms.

Although these mice didn't experience any known side effects from the intervention, breathing thin air around the clock isn't safe for people in the long term. "Given that hypoxia can be dangerous, it should not be attempted at

"When tested in a Parkinson's disease model, the mice improved their coordination and balance"

home," says Mootha. But now, Mootha and his colleagues have combined two experimental drugs to create hypoxia in a pill.

One drug – called osivelotor – alters haemoglobin so it grips oxygen more tightly, preventing the gas from reaching tissues. But this can cause the body to produce a dangerous amount of red blood

cells, which carry oxygen around the body. The second drug, PT2399, prevents this by blocking a protein called HIF-2α, which helps cells respond to low oxygen.

The team gave the pill to mice modelling advanced Leigh syndrome, a severe neurological condition that causes progressive loss of mental and movement abilities. The pill more than doubled the mice's survival time from around 62 days to 158 days.

The researchers also tested the pill in mice modelling Friedreich's ataxia, a rare condition that causes progressive damage to the nervous system, resulting in movement difficulties. The team observed that this stopped further declines in the mice's strength and coordination. Finally, when tested in a Parkinson's disease model, the treated mice improved their coordination and balance

(bioRxiv, doi.org/rhrn). The effects are comparable to when such animals breathe 11 per cent oxygen, says Mootha.

Research suggests that low oxygen can boost brain growth factors, which aid the formation of new connections between neurons. But Mootha proposes that reducing excess oxygen may prevent cellular damage.

The three conditions he and his team focused on are partly characterised by damaged mitochondria, which provide cells with energy. These may be failing to consume oxygen properly, leaving a local excess that harms other cell components, he says.

Gordon Mitchell at the University of Florida says he is "pretty tickled" by the study, but cautions that it remains to be seen how well the effects translate to people. ■

Animal behaviour

Jaguar mysteriously trekked epic 2100 kilometres

A MALE jaguar has made an unprecedented trek, skirting 2100 kilometres of the southern edge of the Amazon rainforest. The journey expands our knowledge of what these big cats are capable of.

The previous dispersal record for a tagged jaguar was about 60 to 100 kilometres, says Jon Morant at University College Dublin in Ireland. He and his colleagues have been tracking jaguars in Brazil via GPS collars as part of an ongoing project that started in 1999. The researchers wanted to better understand how jaguars use their environments along the Araguaia river, a potential wildlife corridor connecting the Cerrado

grasslands with the Amazon.

They captured the male, dubbed Gaspar, as part of this work, fitting him with a collar on 11 November 2024. The 8-year-old jaguar then spent some time in his established

range before setting out on a massive southward trek, skirting the border of the Amazon to inland Cerrado.

The researchers recorded that Gaspar travelled 2122 kilometres,

Gaspar was fitted with a GPS collar – which he took on a surprising adventure

until his collar stopped transmitting locations in late November 2025 (Ecology, doi.org/rhqw). The actual distance is probably much more, says Morant.

While Gaspar mostly stuck to forested areas, he also crossed through heavily modified human landscapes, such as farms, ranches and roads. Dispersals, albeit shorter ones, are common in individuals aged 1 to 3 that are looking to establish a new territory, but it is unclear why an older adult like Gaspar made this move. He might have been searching for better food resources, mating opportunities or a better habitat in general, says Morant. ■ Joshua Rapp Learn

8 August 2026 | New Scientist | 15

---

News

Archaeology

Flattened skulls hint at Iron Age parenting

Christa Lesté-Lasserre

FLAT head syndrome, commonly associated with modern parenting practices, especially placing babies to sleep on their backs, may have been common at least 2800 years ago.

Because humans are born with soft heads, their skulls can get shaped by pressure. Today's use of cots and prams tends to create prolonged pressure on the back of the head, which can deform the normally round surface.

But the earliest history of flat head syndrome – also called positional or developmental plagiocephaly – remains unknown. Thus far, researchers have found two plagiocephalic heads in Pakistan dating from 4000 years ago, as well as examples from 4th-century BC Switzerland and the medieval Czech Republic.

Michael Beck De Lotto at Sapienza University of Rome and his colleagues inspected 26 well-preserved adult skulls from two

“Parents may have placed their babies to sleep on their backs with their heads propped up on pillows”

burial grounds discovered in the 1990s in Matelica, Italy, linked to the Picenes people of the 8th and 9th centuries BC. They found flattened areas on the backs of 11 heads from seven men and four women, making them the earliest known examples of developmental plagiocephaly outside Asia.

The researchers ran CT scans on the skulls and created 2D and 3D virtual models of each one. None of the skulls showed signs of deformation before birth or after death, meaning it must have occurred in the first year of life before head shape solidified, says De Lotto.

Unlike modern flat head

CT scans highlighted where the skulls were flattened (in red)

syndrome – which is mostly concentrated across the back of the head – these Iron Age skulls had a variety of flattening patterns, representing different positions (International Journal of Paleopathology, doi.org/rhrq).

The head shapes suggest parents may have been placing their babies to sleep on their backs with their heads propped up on pillows, he says. They might also have swaddled their babies or placed them in cradles, although the exact mechanisms remain unclear.

The findings align with an existing “evolutionary mismatch” theory about ancient childcare, says Herbert Renz-Polster, a paediatrician and independent researcher in Germany. As hunter-gatherers moved towards a less mobile lifestyle, he explains, they also transitioned from carrying and co-sleeping with their babies to a parenting practice in which babies were placed in bed alone with “sleep aids” like swaddling. That led to babies spending long periods immobile on their backs with extended pressure on their heads. ■

Palaeontology

Dinosaurs were wiped out by global firestorm

James Woodford

THE asteroid strike 66 million years ago caused a planetary-scale firestorm that burned dinosaurs to a crisp all over the world, according to simulations of the impact.

The so-called Chicxulub asteroid landed in what is today the Gulf of Mexico at the end of the Cretaceous Period.

It was thought that most of the worst effects of heat and fires would have been relatively regional, and global extinctions were mainly caused by a “nuclear winter”, where dust blocks out the sun causing dramatic cooling.

Brandon Johnson at Purdue University, Indiana, and his colleagues analysed geological observations and impact simulations, which showed around five times more material was vaporised by the strike than previously thought.

The new work suggests that in the aftermath of the impact, a plume of rock vapour was blasted tens of thousands of kilometres into space and ended up being distributed high above the entirety of the planet.

Just seconds later, the vapour

The asteroid strike sparked a worldwide blast of blistering heat

plume expanded above much of Earth's atmosphere into space, says Johnson, leading to both dust and glassy, metallic beads called spherules, condensing from the vapour.

This immense vapour cloud ultimately turned into 1.6 trillion tonnes of dust, which blanketed the planet, sealing in heat, the study found. The heavier spherules slammed back into the atmosphere (Journal of Geophysical Research Biogeosciences, doi.org/rhq47).

“As this material re-enters the atmosphere over the coming hours, friction decelerates and heats the spherules, creating an approximately hour-long pulse of thermal radiation,” says Johnson.

“This would kill off creatures that were unable to shelter from it and was intense enough to ignite forest litter like grass and pine needles and may have even been enough to ignite trees.”

The geologic record suggests the ensuing fires may have lasted for years, he says. “The Cretaceous animals would have received 17 times the dose of thermal radiation that is 100 per cent lethal to humans,” says Johnson. “That should be enough to kill off even thick-skinned dinosaurs.” ■

SHUTTERSTOCK/KAWINS.COM

16 | New Scientist | 8 August 2026

---

WE WOULD LIKE TO INTRODUCE YOU TO THE WHOLE FAMILY

You are in safe hands with Saga. With more than 70 years' experience of designing and delivering award-winning products for people over 50, you can trust us to provide high quality, value and excellent service.

CRUISES

saga.co.uk/cruises

HOLIDAYS

saga.co.uk/holidays

INSURANCE

saga.co.uk/insurance

MAGAZINE

saga.co.uk/magazine

MONEY

saga.co.uk/money

Find out more about our full range of products at saga.co.uk

SAGA

Experience is everything

NHA-AP6671

---

Views Columnist

Food for thought More and more research is linking ultra-processed foods to depression, anxiety and even dementia. Helen Thomson tucks into the evidence

Helen Thomson is the author of books including Unthinkable: An extraordinary journey through the world's strangest brains and This Book Could Fix Your Life. She was previously head of features at New Scientist

Helen's week

What I'm reading

The Seven Husbands of Evelyn Hugo by Taylor Jenkins Reid – the perfect soon-to-be-poolside read

What I'm watching

Emma Barnett: Fighting Endometriosis

What I'm working on

Season 2 of Change Your Mind with New Scientist, the fab new podcast I co-host with New Scientist's editor Cat de Lange

BY NOW, we all know that ultra-processed foods aren't particularly good for our bodies. Often high in sugar, salt, saturated fat and additives, UPFs have been linked with an increased risk of heart disease, diabetes and some cancers. But what do they do to our brains?

Lately, I've noticed an increasing number of studies linking UPFs with depression, anxiety and even dementia. I wanted to look at how strong the evidence is, and whether it offered another persuasive reason to eliminate them from my diet.

The immediate problem is that, while many observational studies have found associations between greater intake of UPFs and various brain health problems, direct causal evidence is less forthcoming. The picture is further complicated by how we define UPFs, says Tim Spector at Kings College London, co-founder of nutrition company Zoe. Most countries use the NOVA classification system, which is designed to categorise levels of processing, but not to assess how a food affects health.

That said, evidence is mounting for the harms of UPFs. In 2024, a study involving around 30,000 people who were followed for up to 18 years showed that for every 10 per cent increase in the proportion

“Eating more fibre and oily fish are the two best changes you can make for your brain health”

of UPFs in a person's diet, the risk of cognitive impairment and stroke rose significantly. This association remained even when considering factors such as income, alcohol consumption and physical activity.

Another study followed 10,775 adults in Brazil for around eight years. Those who obtained more than 19.9 per cent of their energy

Many ultra-processed foods are high in salt, fat and sugar, which can disrupt the gut microbiome

from UPFs experienced faster declines in global cognition – an overall measure of mental ability – and executive function, which is the ability to control impulses, regulate emotions and stay focused.

It's not just general cognitive decline, either. In 2025, researchers analysing data from 58,423 people between the ages of 40 and 70 found that increased UPF consumption was associated with a higher risk of dementia, Parkinson's disease and multiple sclerosis.

Other studies have linked UPFs to depression and anxiety, too. In one, researchers followed 31,712 female nurses between 2003 and 2017, who didn't have depression at the start of the study. Those eating more than 8.8 servings a day of UPFs – like artificially sweetened drinks or ready-made meals – had a 49 per cent higher risk of developing depression. Although observational, the study used some advanced statistics to rule out “reverse causality” – the idea that if you're depressed, you may be more tempted to eat less healthy foods.

So what might UPFs be doing that is so bad for the brain? The answer is likely to involve multiple,

overlapping pathways. For instance, many UPFs are high in salt, fat and sugar, all of which are known to disrupt the gut microbiome, shifting it so that it has less diversity and fewer beneficial bacteria. This can damage the gut lining, making it “leakier” and allowing harmful substances into the blood and brain. “High-fat and high-sugar diets also induce low-grade inflammation,” says Spector. “We now know that there's a close link between inflammation and mental health, particularly depression.”

Laura Bojarskaite at the University of Oslo in Norway points me to a small study published last year that found that just five days of eating an extra 1500 calories from ultra-processed snacks altered men's brain responses to insulin, a hormone that helps cells take up glucose for energy. The effect remained detectable a week after they returned to their normal diets.

Insulin regulation is vital for maintaining cognitive function, while insulin resistance is linked with depression and Alzheimer's. Although the study didn't look at brain health specifically, it provides evidence that UPFs can rapidly modify brain signalling, which Bojarskaite says is a “credible pathway through which a lifelong [UPF-heavy] diet could influence brain ageing”.

Bojarskaite says that rather than trying to eliminate every UPF from our diet, we should concentrate on reducing products that combine multiple unfavourable features – like added sugars, high salt and low fibre – and replace these first.

If that still sounds too hard, Ted Dinan at University College Cork in Ireland offered a simpler starting point on a recent episode of the podcast I co-host, Change Your Mind with New Scientist: eating more fibre and oily fish are the two best changes you can make for your brain health, he says. ■

18 | New Scientist | 8 August 2026

---

Clumpy cosmos The cosmological principle says everything, everywhere should be roughly the same across the universe. Is it time to give up on that idea, asks Leah Crane

IN COSMOLOGY, there is one principle to rule them all: the aptly named cosmological principle. It insists that there are no special places in the cosmos – on large-enough scales, everything is roughly the same in all directions, from all perspectives. This seems like it must be false: obviously some places in the universe are different from other places. My office, for instance, bears no resemblance to a black hole.

For it to be useful, we must zoom out – which is why modern statements of the cosmological principle tack on “at sufficiently large scales”. Any supercluster of galaxies is just about the same as any other one. It is a cornerstone of our best model of the cosmos, called the standard model.

Even so, a multitude of cosmologists have called this fundamental rule into question.

There are two parts to this principle, homogeneity (everything is smooth rather than lumpy, on large enough scales)

Even the Lobster Nebula looks smooth when you zoom out far enough

and isotropy (that’s true in every direction). The former faces more credible threats, which generally come in the form of structures of unusual size, leading to names such as the Huge Large Quasar Group or the Giant Arc, all of which are billions of light years across. It’s tough to consider our universe to be smooth when it seems to have such enormous lumps in it. The “at sufficiently large scales” that gets inserted in the cosmological principle can only stretch so far, right?

In the early 2000s, astronomers started finding structures that were just a little bit too big to fit within the standard model. Not everyone agrees on what the limit ought to be, but the proposals are generally a couple hundred megaparsecs, or several hundred million light years – anything bigger than that is a problem, and there do seem to be structures that are bigger. So, homogeneity across the universe is questionable.

Personally, I’m not all that worried about the potential breakage of the cosmological principle itself. Some researchers say a little bit of clumpiness is

actually OK within the standard model, some say the universe has expanded so much since its formation that we can’t actually expect the cosmological principle to hold anymore, and some simply say we have to look at an even larger scale. But if I have to zoom out to such a degree to make a paradigm fit the real universe, is that still a useful paradigm?

The cosmological principle lets us make one huge and necessary assumption: that where we are isn’t special, that what we observe from Earth is true everywhere. If we want to crack on with studying the entire universe, we have to assume some level of homogeneity and isotropy. But just because an assumption is useful that doesn’t make it true.

In science, we assume things all the time that aren’t necessarily true, largely to make calculations easier. Every distance we measure to any object in space (and even the distance scale for the cosmological principle itself) assumes a value for the expansion rate of the universe, even though we don’t quite know what it is. We can accept something as close enough to the truth to be useful, while knowing it’s not quite all the way correct.

So, do I believe in the cosmological principle? No, not really. The cosmos is too huge and too full of variety for me to completely buy into the idea that there are no unique places. It’s infinite enough, at least, that the cosmological principle can be both true enough and untrue. There is life in our corner of the universe, so there is probably life elsewhere as well, because in some ways, the universe is the same everywhere. But it is also different everywhere. There are special places, and we are in one of them, and only by holding both that and the cosmological principle in our heads at once can we truly understand our place in the universe. ■

Leah Crane is a senior reporter at New Scientist, covering topics ranging from private space flight and cosmology to quantum mechanics

Leah’s week

What I’m reading The Rime of the Ancient Mariner by Samuel Taylor Coleridge

What I’m watching A whole lot of leopardy reruns, for some reason

What I’m working on Preparing to go to the Arctic to see a total solar eclipse!

LIGHT, BIRARD, MALAYSIA, CALTECH

8 August 2026 | New Scientist | 19

---

Views Columnist

Drugs for 'undruggable' proteins A new type of drug that works in a completely different way has been approved, and it could revolutionise treatment options, says Michael Le Page

Michael Le Page is a reporter at New Scientist specialising in genetics, biomedicine and the environment

Michael's week

What I'm reading I just finished Atmosphere by Taylor Jenkins Reid

What I'm watching I'm finally getting around to watching The Dinosaurs on Netflix

What I'm working on I'm already interviewing people for what might be my next column, about muscles

I THINK we will look back at 2026 as a landmark in medicine, the year in which the first protac drug was approved. While most drugs simply block their targets, protacs utterly demolish them. Per molecule, protacs are far more potent, and they can also do things other drugs just can't.

To understand why protacs are so revolutionary, we need to start with conventional drugs. Almost all the pills you pop contain drugs that consist of small molecules. The shape of these molecules allows them to bind to specific proteins, to block or stimulate them.

Small-molecule drugs have some big advantages. They're cheap to make, can get inside cells and can be taken as a pill because they can also pass through the gut lining and into the bloodstream. But their small size means they're not that specific. They tend to bind to other molecules besides the one they are supposed to target and thus have undesirable side effects.

And in many cases, it simply isn't possible to find a small molecule to block the activity of a protein. The issue is that each drug molecule can't just bind to any part of a target protein. To have an effect, it must bind to the active site of the protein – the workface – but

"In trials, people with advanced breast cancer survived more than twice as long when given the protac drug"

sometimes small parts of the active site don't have a distinctive enough shape. Such proteins are referred to as undruggable.

That's why many newer drugs consist of large molecules, most commonly proteins known as antibodies. (Any drug whose name ends in "-ab" is an antibody.) The large size of antibodies means they

The US Food and Drug Administration approved a protac drug for breast cancer in May

can bind to targets with exquisite precision, making them highly effective and safer than small-molecule drugs. To create them, you get immune cells to generate antibodies to the desired target and then pick the best one. It sounds easy, but it took decades to perfect.

The downside of antibodies is that they are really expensive because they can be made only in living cells. Their large size also means they must be injected because they can't get through the gut. Nor can they get into cells once they're in the bloodstream. So, while antibodies could bind to many undruggable proteins, they can't get into cells to reach them.

The other thing to mention is that both small-molecule drugs and antibodies must be in contact with the target to work. Protacs work differently. Essentially, they stick a "recycle me" label on a protein in a cell. That means just one protac molecule can label and get rid of tens or hundreds of molecules of that protein. And because they're not blocking or stimulating the target protein, they don't have to bind to the active site. Anywhere on the target will do, making many undruggable targets druggable.

Creating protac molecules that were small enough to be useful was a challenge, because each protac must bind to two molecules, whereas conventional small-molecule drugs need bind to only one. Craig Crews, who came up with the idea for protacs with Raymond Deshaies in 1998, succeeded in making them small enough to behave like small molecules, but they are still somewhat larger than conventional drugs. "I call them large small-molecules," Crews told me last month.

By 2013, Crews was confident enough to found a company called Arvinas. For its first products, Arvinas played it safe. Rather than going for undruggable proteins, it targeted hormone receptors in cancer cells. For instance, the growth of some types of breast cancer is driven by receptors that respond to oestrogen. There are already several drugs that work by targeting these receptors, but resistance can develop.

One way that cancers become resistant is by making more receptors, until even the highest tolerable drug doses can't block all the receptor molecules. "Since we're actively degrading the receptor, we don't have that limitation," Crews told me in 2020. In trials, people with advanced breast cancer that had spread around the body survived more than twice as long when given the protac drug vepdegestrant, compared with those given a conventional drug. As a result, vepdegestrant was approved by the US Food and Drug Administration in May.

"The approval is a milestone for the field," Crews told me last month.

I'm hopeful we will soon be seeing all kinds of new treatment based on protacs, ultimately helping billions of us. The approval of vepdegestrant may have received barely a headline, but if it is indeed the first of many, we will look back at it as a major milestone. ■

20 | New Scientist | 8 August 2026

---

New Scientist Subscriptions

HALF-PRICE SALE ONE YEAR FOR £99

Unleash your curiosity this summer and explore the latest scientific ideas anytime, anywhere with unlimited digital access.

Subscribe today at newscientist.com/summersale

To order by phone call +44 (0) 330 333 9470, quoting 22393. These are auto-renewing subscriptions that can be cancelled at any time. We offer a 14-day cooling-off period after the initial payment is made. Offer ends 06 October 2026.

---

Views Aperture

22 | New Scientist | 8 August 2026

---

Net loss

Gaurav Patil Mangrove Photography Awards

THIS unusual photograph, shot under ultraviolet light with a 13-second exposure, shows some of the damage caused on land by fishing out at sea.

Gaurav Patil's image of a fishing net that has been carried by ocean currents and become tangled in a mangrove tree on the shore of Nandgaon village, south of Mumbai, India, has won the overall prize in the 2026 Mangrove Photography Awards. The competition is organised by US non-profit organisation Mangrove Action Project to raise awareness of the precarious position of the world's mangrove forests.

Mangroves make up less than 1 per cent of all tropical forests, but provide crucial habitat for wildlife and store up to 10 times more carbon than the same area of inland forest. Despite better recent news after decades of decline, more than half of mangrove ecosystems are thought to be at risk of collapse by 2050.

Hundreds of thousands of tonnes of lost, abandoned and discarded fishing gear litter the oceans, carried around the globe by currents and eventually getting caught up or breaking down into microplastics. This gear can severely damage mangrove ecosystems, entangling marine life and choking mangrove trees.

"This photographer has put life into a subject that normally looks dull but is deeply important for conservation. Ghost nets are a real problem," said Judge Dhritiman Mukherjee in an announcement about the prize. ■

Matthew Sparkes

8 August 2026 | New Scientist | 23

---

Views Book of the week

The trouble with life

A provocative book sets out to describe the living world very differently – by insisting on agency and purpose everywhere. Michael Marshall explores

Book

Alive

Melanie Challenger

Canongate (UK);

Penguin Random House (US)

WHEN you think about it, being alive is a really strange state of being. The vast majority of the matter in the universe is lifeless hydrogen and helium, rocks and dust. Yet, somehow, here we all are: millions of species of microbes and plants and animals, all respiring and eating and trying to reproduce, some of us squawking, and some sleeping, all of us eventually dying. When we consider the sheer implausibility of life in the cosmos, wrote biochemist Jacques Monod in Chance and Necessity, “we should feel strange and a little unreal”.

Actually defining what life is has proved extremely tricky, so we tend to fall back on metaphors. One of the most prominent of the past century or so is that living organisms are machines. They have components, moving parts, complicated chemical signalling

pathways to make things happen. In this view, even the brain is a kind of computer, one that responds to stimuli and performs calculations. Moreover, the machines have a purpose, shaped by evolution. In The Selfish Gene, Richard Dawkins argued that organisms are machines that exist to pass on their genes, that living things are just complicated by-products of genes copying themselves.

In Alive: The hidden intelligence of the living world, writer and broadcaster Melanie Challenger sets out to completely demolish this line of thinking. Living beings aren’t machines, she argues, nor do they exist purely to pass on genes. Instead, what defines living beings is agency and purpose. “Life-forms pursue goals because they are alive, and life is, by its nature, purposeful,” she writes.

The idea that organisms have goals and purposes is “the most obvious” answer to what life is, and also “the most curiously neglected”, argues Challenger.

From René Descartes onwards, the idea that non-human animals have goals, feelings and even thoughts of their own has been

treated with scepticism by zoologists and psychologists. Animal behaviourist C. Lloyd Morgan famously argued that we should never interpret an animal’s behaviour as evidence of a higher psychological process if it could also be explained using a simpler process. “Morgan’s Canon” went on to guide a century of research – even though it isn’t obvious what is meant by higher and lower psychological processes.

Challenger’s suggestion is that we can’t escape the idea of a self, something with agency, to explain the behaviours of living organisms. It’s just that we have removed it from the organisms themselves: either in the form of some separable soul or essence, or a set of genes that determine how organisms behave for their own ends. “Both soul and gene are described as self-like agents, the true initiators of the acts of a life-form,” she writes. In these views of the world, “life-forms themselves are marginalised”.

Instead, we should see living beings as “self-determining”, she argues: “Life-forms use energy to remain locally discontinuous with

the environment... and what is phenomenal about this act is that it is self-generated.” Challenger develops this idea over six chapters, exploring homeostasis (the mechanisms by which organisms maintain their internal environment), sensory perception, consciousness and intelligence. These can all be explained mechanistically using cells, molecules and physical forces, but they only make sense in the light of the whole organism.

Challenger is trying to reclaim the active nature of life. Even organisms like plants that don’t have brains, and therefore may have little or no conscious experience, are still responsive to their environments and act with purpose for their own ends. There is nothing automaton-like about animals or plants: that’s why the automata of sci-fi are so deeply, gut-level disturbing in contrast.

Cards on the table: I think

Three more great reads on the meaning of life

Life’s Edge: The search for what it means to be alive by Carl Zimmer

Despite hundreds of attempts, we still don’t have an agreed definition of life. Science writer Carl Zimmer explores the many tricky edge cases, like viruses, which blur the line between life and non-life.

Togetherness: Symbiosis and the hidden story of life’s greatest collaborations by Rowan Hooper

An ideal companion piece to Alive, Rowan Hooper (New Scientist’s podcast editor) shows us the importance of understanding life as a community. Individual organisms are only part of the story.

Free Agents: How evolution gave us free will by Kevin Mitchell

Neuroscientist Kevin Mitchell argues that living beings are “agents acting with purpose”, not mere machines responding to stimuli. His ideas were among the inspirations for Melanie Challenger.

24 | New Scientist | 8 August 2026

---

Challenger is mostly right. The gene's-eye view of evolution misleads us because it gives too much agency and power to the genes. In reality, genes aren't master controllers of cells, nor do they contain blueprints for entire organisms: they are part of a web of interacting systems that underpin life. Genes are important in cells, but are no more in control than any other component.

I can imagine two readerships for Alive, with drastically different responses. One valid reaction is that this is blindingly obvious. It's transparently clear that animals act with purpose, out of sheer vitality and hunger, and joy and will to live. The thing is, the idea of life-as-automata has acquired a lot of power. Challenger argues that it developed alongside the rise of industrialised farming, which exploits and mistreats animals on a staggering scale: if you don't quite believe animals

are sentient, it's easier to accept this globalised abuse. Today, synthetic biologists often treat life as, literally, programmable matter: they are the second readership, who will probably find Challenger's position difficult to accept.

I do think Challenger has one major difficulty: her argument focuses on individual organisms. Whether it's a bacterium or a shark, each acts for its own purposes. She

A Brazilian tropical forest supports trillions of individual organisms

is right to stress that individuals have great agency, but finds herself having to neglect or downplay the fact that individuality is fuzzy, and that individuals can only survive in ecosystems.

Consider your body. Yes, you are you and you act for yourself. But you are also home to trillions of microbes, which influence you, and some of which you couldn't survive without. Challenger insists on treating the microbes as distinct from you; each follows its own goals and it just so happens they mostly coincide with yours. But this is, I think, to oversimplify.

She has a similar problem with eusocial insects, in which a single individual reproduces and the rest of the nest is sterile. Eusociality is one of those phenomena that does need explaining at the level of genes, at least to some extent. Challenger doesn't like labelling eusocial ant colonies as "superorganisms", but her justification is weak. She argues that worker ants are still capable of rebelling against the system, so they aren't fully integrated, but the same could be said of some of our cells – that's why we get cancer.

My one other caveat is that her high-flown metaphors and philosophy-speak won't be for everyone. Here's an example: "The activities of organisms are the literal embodiment of what we take the features of our world to mean, radically determined by the organisms themselves." If you have a low tolerance for this, you may find the book tricky, but the effort is worth it: Alive will help you see the living world afresh. ■

Michael Marshall is a science writer based in Devon, UK

New Scientist recommends

Eleanor Parsons Magazine editor London

Peering at the fossilised skeleton, I see something extraordinary. In among the 182-million-year-old ichthyosaur's ribs are the remains of another, smaller reptile, probably another ichthyosaur –

a cannibalistic dinner, or a fossilised pregnancy. This specimen is on display as part of Jurassic Oceans: Monsters of the deep at the Natural History Museum in London until 3 January. The exhibition uses fossils, casts and video to tell the story of some of the fiercest ocean predators that ever lived.

There are the large jaws from a crocodile-like short-necked pliosaur. You can even touch the tooth from a mosasaur (its skull is pictured above), which lived a little later, in the Cretaceous. It is pointy but smooth, with gentle ridges indicating growth lines, if you are wondering how it feels.

I loved the ichthyosaur eye and the fish bone-studded poo from an ichthyosaur or plesiosaur. But my favourite was the fossilised angel shark, looking as if it could still be lying on the seabed, waiting for its prey.

THE TRUSTEES OF THE NATURAL HISTORY MUSEUM LONDON

8 August 2026 | New Scientist | 25

---

Views Culture

Into the unknown

An in-depth history of the successes and failures of space exploration reminds us of what faces all who leave our planet, finds Karmela Padavic-Callaghan

Book

Out of This World

Peter Bond

Reaktion (UK);

University of Chicago Press (US)

DO RUSSIAN cosmonauts and US astronauts don't their spacesuits differently? What was the first musical instrument played on the International Space Station (ISS)? And who was the first person to smuggle pickles into orbit?

In 1957, the Soviet Union launched Sputnik 1, the first artificial satellite. Since then, humanity has explored space by simply going there, a feat replete with challenges, tragedies and oddities. In Out of This World: Life as an astronaut, writer Peter Bond chronicles many of these in great detail, offering a well-informed and unfussy view of space travel and life in space alike.

His book is thematic rather than chronological, with the first chapter aptly titled 'Learning to Leave Earth'. It's an overview of early space flights and the development of the rockets and aircraft that made them possible, such as the second world war V-2 rocket – the foundation for the first generation of space launchers.

Space exploration, shows Bond, has always been linked to large nations jockeying for power. This section could have been enriched by more anecdotes and personal details from the pioneers. Instead, Out of This World moves quickly and covers a remarkable amount of ground. Bond's next focus is the process by which different nations' space agencies, such as NASA, select their astronauts, and how strenuously they must train.

Russian cosmonauts live and train in Star City, their own isolated townlet north-east of Moscow, and

NASA's recruits primarily undergo training in Houston, Texas. If someone has flown in space, you will probably find their name somewhere in Out of This World.

We also hear from the astronauts themselves as Bond quotes their

'The Apollo 7 crew got into a fight with mission control over wearing their helmets when they had colds'

diaries and interviews, providing his book with the necessary sparks of humanity. I was touched to read about the time astronaut Michael Collins spent in the Apollo 11 command module while the rest of the crew walked on the moon for the first time. 'Not since Adam has man known such solitude as Mike Collins when he passes around the back of the

Moon,' he wrote in the mission log.

Astronauts also get grumpy on the job just like everyone else. For instance, Bond reveals that in 1968, the crew of Apollo 7 got into a fight with mission control over wearing their helmets while their noses were stuffy from a cold inadvertently brought on board.

While Bond primarily discusses US and Russian missions to space, China's taikonauts often get the spotlight, too, perhaps pointing towards a more international, though certainly no less politically fraught, future. These tensions are prominent in Bond's penultimate chapter, which looks at lunar missions. The cold war race to put a man on the moon is an example of the US outdoing the Soviet Union, but space exploration never became politically simple. US efforts to return people to the moon exemplify this.

Clockwise from left:

Artemis II's launch in 2026; the Apollo 13 crew in 1970; Valentina Tereshkova, the first woman in space; Bruce McCandless II's 1984 untethered spacewalk

At the time of writing, Bond acknowledges that there are many questions hanging over NASA's Artemis mission, which is meant to accomplish the task. By the time I read Out of This World, part of the mission had been triumphantly completed, and part did, in fact, suffer restructuring. Notably, Bond has little to no discussion of debates surrounding ideas of colonising the moon and Mars for private and national profit.

The book ends with a solemn recounting of all the lives lost in space, from well-known incidents such as the Challenger disaster in 1986, which killed seven crew, to

26 | New Scientist | 8 August 2026

---

Rewinding ancient history

Tom Hiddleston is the unlikely host of a Pompeii documentary. It is unabashedly cheesy, but somehow it works, says Josh Bell

fatalities among early Russian cosmonauts. Bond's writing is never lyrical, but a little more emotion creeps into this last chapter: he is writing to honour the lost lives, and explain how their superiors and support teams failed them. If 350 pages in which he patiently unpacks the many ways space exploration is remarkably challenging aren't enough to curb remaining human hubris, a reminder of how fragile and perishable we are in the face of cosmic vastness might just work.

The next crewed flight to the ISS is planned for next month. I will be awed and anxious as I think about it – and the long history of space flight captured in Out of This World. Thanks to Bond, I'll also remember delightful stories like that of Russian cosmonaut Igor Volk, who smuggled pickles onto the Soyuz T-12 mission in July 1984. ■

Television Pompeii: Out of Time Disney+

WHEN the viral trend of asking men how often they think about the Roman Empire was sweeping social media in 2023, it was clearly a mistake that no one posed the question to actor Tom Hiddleston. He apparently thinks about the Roman Empire so often that he has now produced and hosted an entire three-episode National Geographic docuseries about it.

Early in the first episode of Pompeii: Out of Time, Hiddleston recounts his formative visit to the preserved remains of the ancient Roman city in 1998 when he was a teenager, showcasing some charming photos of his younger self on the kind of trip where others might spend their time drinking and partying. Instead, Hiddleston says that his experience at Pompeii shaped him so profoundly that it led him to study classics and Latin at university, and even to his decision to become an actor.

While he consults multiple scholars in Out of Time, the show plays up Hiddleston's abilities as a performer to provide insights into the people who lived in the towns that were destroyed by the eruption of Mount Vesuvius in AD 79. "I need to combine the historical evidence with what I do as an actor," he says, with all the solemnity of someone who takes his craft very, very seriously.

Hiddleston's unabashed cheesiness is what makes the show work, though. Before introducing his first expert, he touts revolutionary new technology that allows for

the tracking of actual residents of Pompeii, only for it turn out to be a simple spreadsheet created by classics professor Steven L. Tuck. Hiddleston then conjures up the special effects that turn Tuck's data into a three-dimensional holograph, noting cheekily: "I've seen some Avengers movies."

His experience playing Loki in the Marvel Cinematic Universe continues to inform Out of Time, which counts former Marvel Studios executive Kevin R. Wright among its producers. Hiddleston channels Loki with a little gesture of his thumb and forefinger, deployed multiple times throughout the series to pause or rewind moments when incorporating new information. The series mixes narration and interviews with dramatic re-enactments, and Hiddleston often appears right alongside the actors playing ancient Romans, like Loki travelling through time to witness historical events.

Those dramatisations feature better acting than typical historical recreations, and they are more grounded than director Paul W.S. Anderson's bombastic 2014 action movie Pompeii, the most recent high-profile

Geologist Chris Jackson (left) and Tom Hiddleston meet on Mount Vesuvius

pop-culture take on this topic. Director Tom Barbor-Might and the show's writers create the emotional engagement that Hiddleston is aiming for when he picks out three specific real-life people to focus on: teenage blacksmith's apprentice Avianus (Frankie Treadaway), wealthy businesswoman Julia Felix (Lubna Azabal) and a retired soldier, the Praetorian (Barry Aird).

Initially, Hiddleston teases the idea that this approach will lead to new revelations about what actually happened, but of course that is no truer than anything Loki the trickster god has ever said. Still, that doesn't make Out of Time any less entertaining or informative, especially for viewers who don't think about the Roman Empire as frequently as those men on social media. Utilising expertise from historians, archaeologists, geoscientists and one "disaster psychologist", Out of Time lays out the scientific and cultural contexts of the events in Pompeii.

Hiddleston does a lot of pensive staring into the middle distance, and his clearly scripted banter with the experts occasionally sounds a bit stilted. It's hard not to laugh when he first unveils his corkboard full of pinned documents, like he is conducting a murder investigation. It's even sillier when his "office" shakes and buckles with the rumblings of the earthquakes in Pompeii.

But Hiddleston also demonstrates genuine, infectious enthusiasm for the subject, and that goes a long way. If he has to make a few Marvel references to get people's attention, that's a small price to pay. ■

Josh Bell is a writer and film critic based in Las Vegas, US

8 August 2026 | New Scientist | 27

---

Save up to 47% on second release tickets

Second release

ENDS

13 AUGUST

NewScientist Live

BOOK TODAY,

PAY LESS

4 STAGES 50+ SPEAKERS

INTERACTIVE EXPERIENCES AROUND THE SHOW

10 – 11 October

Schools' Day 12 October

EXCEL LONDON

• ONLINE

SECURE YOUR TICKETS AT THE BEST AVAILABLE PRICE

newscientist.com/nsl2026

PROUDLY SUPPORTED BY

Yakult

THE HUSON CLUSTER

FARMERS

KING'S College LONDON

Medical Research Foundation

Technicians We make the difference

THE PAPER

47% saving applies to weekend family tickets, other ticket options also available. If you are a paid subscriber to New Scientist, enter your 12-digit subscriber number (including any zeros) at checkout for 10% off the current ticket price and 50% off an online upgrade.

---

Views Your letters

Editor's pick

Restricting open-source software won't stop misuse

Leader, 25 July

From Sam Edge,

Ringwood, Hampshire, UK

You seem to criticise open-source software on the grounds that it can give everyone access to potentially dangerous technology. However, there is a saying in cryptography that applies equally well here: "Security through obscurity is no security at all." Just as trying to restrict access to books on chemistry and ballistics doesn't stop rogue actors from killing people in large numbers, neither will trying to ban open-source control systems or large language models. In any case, closed-source software can also be reverse engineered, its source code stolen via industrial espionage or simply pirated and used for off-brand purposes.

How to acclimatise to heat, courtesy of a Floridian

18 July, p 16

From Craig MacDougal,

Tampa, Florida, US

The conditions Alec Luhn experienced inside a heat chamber are considered normal for summer afternoons where I live in Florida. Here's a tip from a subtropical climate: to get a sense of how miserable it is going to be on a given day, focus on the dew point, not the relative humidity. The latter starts high in the morning and goes down as the temperature rises, possibly giving the impression that it will feel better in the afternoon. (It won't.)

The dew point – the temperature where the relative humidity is 100 per cent – however, will stay relatively constant. Once it rises to about 21°C (70°F), it feels humid, regardless of the air temperature. I calculated the dew point for Alec's test as about 23°C (73°F). A dew point of 25°C (77°F) or more, which happens around here, is miserable and can be dangerous.

I moved to Tampa from Jacksonville in north-east Florida in 1978. I am acclimated. At age 69, I play tennis twice a week, drink my sports drink and sweat a lot. Not everybody can do that, but I will for as long as I am able.

Shackleton photo hints at a forgotten injustice

18 July, p 18

From Margaret Cuthbert,

Edinburgh, UK

You recently published a dramatic photo of the James Caird, the boat that carried Ernest Shackleton and five of his crew across the Antarctic waters to South Georgia in 1916. That image also highlights a long-standing injustice.

Following the loss of his ship, the Endurance, Shackleton had planned for the crew to haul their three lifeboats over the ice and snow until they reached open water. This was met with dismay by carpenter Harry McNish, who foresaw that serious damage to the boats from being pulled over rough terrain was likely to occur. The row ended with Shackleton reportedly levelling a gun at McNish and winning the argument, but within two days, he had to abandon his plans and return to their base.

Eventually, when the ice broke up, they sailed their three lifeboats to Elephant Island. It was there that McNish modified the largest of the three, the James Caird, and made it seaworthy for the journey to South Georgia. The photograph you published gives some indication of how extensive McNish's modifications were. This work, carried out in freezing cold and with frostbitten fingers, was a vital contribution to the survival of the explorers, and only McNish had the skills to make it.

However, when Polar Medals were later handed out, McNish was deprived of his due, after Shackleton failed to recommend him for one. This is an injustice that it is not too late to rectify.

Intelligent life wouldn't be power-hungry like us

4 July, p 41

From Dyane Sylvester,

Arnside, Cumbria, UK

Brian Lacki shows a typically anthropomorphic bias in his search for extraterrestrial intelligence by assuming that any "advanced civilisation" will have developed computers and require huge amounts of power. Wouldn't a truly intelligent species be using simple technology requiring minimal power and resources, to ensure that it could exist within the capacity of its home planet?

It's hard to tally economics with the environment

Letters, 18 July

From Tim Moulsley,

Caterham, Surrey, UK

The expenditure by the Australian government to protect the Great Barrier Reef, as described in Keith Noble's letter, is no doubt to be applauded. However, one should take care before concluding that this is a very good thing if that is based on the reef supporting 77,000 jobs and making an annual contribution of AUS$9 billion to the country's economy, as Keith mentions. Unless that economic activity is strictly net zero, it will be directly contributing to the climate change that is the primary factor necessitating protective measures in the first place!

I also wonder how much of this activity could actually be classed as the "useless overconsumption"

Joel Millward-Hopkins mentions in the article "Living the good life" (18 July, p 32), and thus something to be curbed in our efforts to ensure everyone has a reasonable basic quality of life, while keeping within the planet's safe operating space. I fully appreciate the great appeal of the reef for tourism, but it is hard to see that this will have a long-term future, unless, as part of the sum total of all other human activities, the proposed "Great Barrier Reef economic zone" becomes truly sustainable.

Don't forget balance when it comes to brain health

11 July, p 32

From Wai Wong,

Melbourne, Australia

An often neglected measure of brain health is balance. While exercise in general is good for brain health, a regime that targets balance, strength and movement, such as tai chi, improves cognitive function better than approaches focusing on just one area. Various forms of simple and free balance tests are available from trusted sites on the internet. So, it pays to include balance in your brain check-up and exercise routine.

Another option for geothermal in the UK

25 July, p 32

From Larry Stoter,

The North, Monmouthshire, UK Your article on new geothermal methods makes no mention of using abandoned mines. In the UK, there are many such ex-coal mines that could be used as a heat source. While temperatures may not be sufficiently high to generate electricity, they are high enough for district heating schemes. Such schemes are already in action at Gateshead.

For the record

Iceland's Blue Lagoon is an artificial geothermal spa (25 July, p 32).

Want to get in touch?

Send letters to letters@newscientist.com;

see terms at newscientist.com/letters

Letters sent to New Scientist, 9 Derry Street, London, W8 5HY will be delayed

8 August 2026 | New Scientist | 29

---

Features Cover story

ANDREA DE SANTIS

30 | New Scientist | 8 August 2026

---

The prediction machine

A new kind of artificial intelligence is starting to understand cause and effect in the physical world, discovers Daniel Cossins

WHAT happens when you knock a glass off a table? Ask ChatGPT and you get a step-by-step explanation: how gravity accelerates the receptacle along a “parabolic trajectory”, why it could shatter “if the impact generates stresses greater than the material can withstand”. Ask my 3-year-old daughter and she’s less bothered with fine details: “It go smash and water everywhere!”

When it comes to the future of artificial intelligence, the difference is instructive. The large language models (LLMs) powering today’s chatbots are trained on vast swathes of text to predict the next word in a sequence. Unlike little Adeline, however, no LLM has ever actually knocked a glass off a table – or thrown spaghetti at a wall, or driven a scooter into a pond – and that may impose a fundamental limitation on their capabilities.

This, essentially, is why some researchers argue that what we need to take AI to the next level isn’t ever-larger LLMs, but “world models”: systems that learn through observation so they can simulate the consequences of actions in the real world.

World models have rapidly become AI’s next frontier, attracting huge buzz from industry and business thanks to their promise in autonomous robotics. What makes the world-model approach especially intriguing, though, is that it also ostensibly offers a route to artificial general intelligence (AGI), or machines with human-level reasoning that can be applied across a range of tasks.

The problem is that it is tricky to parse substance from bluster. The term “world model” is confusingly elastic, for starters – for some, problematically so. It has become “a kind of shorthand for all the things that current AI systems can’t do well”, says Melanie

Mitchell at the Santa Fe Institute in New Mexico. It is unclear exactly what these systems should model, never mind whether internal representations of physical reality will be sufficient for the kind of generalisable intelligence that means my 3-year-old intuitively understands what happens when she swipes a glass, even if she can’t say why.

To make sense of world models as they apply in AI, it helps to understand the concept’s roots in cognitive science. AI researchers themselves typically trace it back to 1943, when psychologist Kenneth Craik wrote that the human mind “carries a small-scale model of external reality and of its own possible actions”, allowing us “to try out various alternatives” and “react to future situations before they arise”.

That idea has evolved in the decades since, most notably with the theory of predictive processing. This posits that perception – possibly even consciousness itself – relies on the brain constantly generating predictions about the external world and updating them in response to incoming sensory data.

But the point remains. “The key idea is

that intelligence involves building models of the world in our heads so we can simulate outcomes and avoid costly mistakes,” says Josh Tenenbaum, a cognitive scientist and AI researcher at the Massachusetts Institute of Technology. “So ‘our ideas die in our stead’, as Craik put it.”

It doesn’t take a genius to see why that has piqued the interest of AI researchers. LLMs have proved astonishingly capable, and they can certainly give the impression they understand things as we do. But language is a description of reality, rather than reality itself, and their lack of direct experience means that even the most powerful language models struggle when it comes to “understanding” physical phenomena in the real world.

LLMs often perform poorly when they are asked to reason about spatial concepts and things that might happen in the physical world, says Mitchell. “If you’ve ever uploaded a map to one of these models and asked it questions about it, you’ll know that they will often have a lot of problems with reasoning.” Indeed, in a 2024 study, when researchers trained language models on a database of turn-by-turn directions for taxi trips around New York City, they found that the system could provide reasonable routes from one point to another – but failed miserably when asked to take the odd detour.

The reason why is that the LLMs have only descriptions of journeys, and therefore lack the sort of mental map that allows us to imagine what would happen in different scenarios. And navigation is just one example. LLMs demonstrate similar shortcomings in any scenario where you need to simulate the physical world – safely loading a dishwasher, say, or folding laundry. “They’re not

“LLMs lack the mental map that allows us to imagine different scenarios”

8 August 2026 | New Scientist | 31

---

Google DeepMind's Genie 3 creates worlds that users can explore from text-based prompts

designed for it, which is why people are interested in alternatives like world models," says Tenenbaum.

Probably the most influential proponent is Yann LeCun, a computer scientist at New York University and, until recently, chief scientist of foundational AI research at Meta. His argument, first made in a 2022 paper, is essentially that if we want to build truly intelligent systems that can reason, plan and act effectively in the real world, we need world models.

"I cannot imagine we can build agentic systems without those systems having an

ability to predict, in advance, what the consequences of their actions are going to be," LeCun told artificial intelligence forum AI House Davos in January. The key to that, he reckons, is systems that learn the rules of the world from observation.

The idea has caught on. In December 2025, LeCun left Meta to found AMI Labs, raising just north of $1 billion to build world-model systems. A year earlier, Fei-Fei Li at Stanford University in California started World Labs, with $230 million of investment, to develop AI with "spatial intelligence". Many of the established AI firms, most notably Google

DeepMind, are now also actively pursuing world models in one form or another.

The reason for the influx of money is primarily the promise that world models hold for advancing robotics (see "What are AI world models good for?", below). However, it is early days and the existing prototypes are modest in their applications. With Marble, for instance, World Labs has built a system that generates coherent 3D scenes from text prompts. Similarly, DeepMind's Genie 3 creates convincing interactive virtual environments – "snowy mountain at dusk", say – in which you can move around for several minutes and even prompt events like rain.

In both cases, the companies describe their systems as "world models". But Marble is really a 3D video generator and Genie 3 a video-game simulator, albeit one producing simulations in which agents could plausibly act, observe consequences and learn.

Arguably, a proper world model would be something that exists inside an agent such that it can predict the consequences of its decisions, imagine the future and plan ahead before taking actions – and that is what DeepMind is pushing towards with its Dreamer 4 system,

What are AI world models good for?

For anyone wondering what the rise of world models in AI means for us, the applications researchers have in mind are illuminating. In short, world models are unlikely to produce another ChatGPT moment, when millions of people suddenly discovered a powerful tool they could use in their everyday lives. The promise isn't better chatbots, or really anything that consumers will notice any time soon, but rather a new generation of AI systems whose earliest impact could come in the form of autonomous robots capable of working in factories and warehouses, before possibly making their way into our homes.

The rationale is straightforward: rather than learning by repeatedly trying and failing for real, which is dangerous, or learning responses for every situation from exhaustive training data, which is expensive, robots controlled by AI world models could predict what would happen across a vast range of scenarios and

choose the safest and most effective action.

"There's huge potential for robotics to change everything," says Danijar Hafner at Google DeepMind in San Francisco, who works on world models.

Beyond robots, world models are also touted as a way to accelerate scientific advances, particularly in materials science and drug discovery, where the challenge

lies in exploring vast numbers of candidate molecules or atomic arrangements for novel compounds. By learning the underlying physics, they could simulate countless possibilities – how molecules interact with proteins, say, or which hypothetical compounds best capture carbon dioxide – before anything is made or tested in the lab.

"The key capability is planning in imagination space because, once you have this abstract predictor, you can say, for example, if you want to develop a drug, 'OK, what if I add this chemical or do this series of actions', and you can actually predict what will happen without having to do it in the real world," says Randall Balestriero at Brown University in Rhode Island. "It just makes things cheaper and faster."

World models may help with humanoid robotics

32 | New Scientist | 8 August 2026

GEORGE DEEPMIND

IMAGE | JAMES GREGG | VIX GETTY IMAGES

---

released in September 2025. It learns an internal predictive model of its environment, training mostly on data from the video game Minecraft, and uses that to train an agent that repeatedly “dreams” the consequences of possible actions to improve its behaviour.

“The biggest difference [from Genie 3] is that Dreamer 4 is an agent, not just a world model,” says Danijar Hafner at DeepMind in San Francisco, who leads the Dreamer 4 team. “It predicts actions and improves them through iterative self-improvement, using planning and [imagined] trial and error.”

Playing games

The power of this approach is apparent in Hafner and his colleagues’ demonstration that Dreamer 4 can figure out how to collect diamonds in Minecraft, a complex task involving thousands of different actions – gathering resources, crafting tools, navigating the landscape – without being shown how to play. “The system has to understand its environment and generalise, because each new episode starts in a randomly generated world,” says Hafner.

That is an important step, not least because this is precisely the kind of world model that could facilitate autonomous robots capable of folding laundry, say, or loading the dishwasher. “I think it’s going to solve robotics,” says Hafner. “What’s missing now is execution: data, compute, scaling. But we have the recipe, so, like with language models, it’s about scaling and details.”

Who would bet against DeepMind, given its impressive track record? After all, its researchers have won a Nobel prize for work on protein folding. But while Dreamer 4 demonstrates that agents with internal world models can reason and plan, it leaves a deeper question unresolved: what exactly should these systems learn about the world, or, more specifically, at what level of detail? This has become a key fault line in the field, with two distinct approaches emerging.

Many of the existing strategies are attempting to generate predictions on condition of action, as the researchers put it, by reconstructing future observations as faithfully to the training data as possible. But LeCun reckons that is the wrong approach, or at least not the best. His argument is based on the fact that humans don’t mentally simulate the world in any great detail, and certainly not pixel by pixel. When we imagine a glass falling

from a table and smashing to pieces, for example, we don’t predict the position of every shard of glass, every water droplet. Instead, LeCun argues, we run highly compressed models that capture only the aspects that matter.

This why LeCun advocates for a different tack, which he calls joint-embedding predictive architecture (JEPA). In this framework, world models infer abstract representations of what is relevant for reasoning, planning and action. “Generative models try to reconstruct pixels, whereas JEPA learns in a latent space and only predicts what is useful,” says Randall Balestriero at Brown University in Rhode Island, who has worked with LeCun on JEPA-based world models. “It ignores irrelevant details and focuses only on what matters for the agent.”

LeCun and his investors appear to be betting that the JEPA approach will offer a swifter route to real-world applications, largely because it requires less training data. The most concrete publicly available demonstration of this technique so far is a system called V-JEPA 2, released by Meta in June 2025. Trained on video inputs, V-JEPA learns by masking certain regions of footage – obscuring a moving ball,

say, over multiple consecutive frames – and repeatedly predicting not pixels, but abstract representations of what was hidden to learn a compact internal model of how the ball moves. What LeCun and his colleagues have shown, then, is that their system can model the causal structure of the physical world without having to reconstruct it in detail.

Which isn’t to say that JEPA is necessarily any better than generative world models. Its advocates argue that it will be, of course. “Without abstraction, AI systems will stay limited to narrow tasks,” says Balestriero. But V-JEPA 2 is yet to clearly demonstrate that its abstract representations are sufficiently meaningful to enable an agent operating in an open-ended environment to reason and plan. “The biggest missing piece is the question of, how do we know that the abstraction is actually useful,” says Balestriero. “This is where there is a huge amount of active research right now, to understand: what do you capture, or how do you encode something very rich about the world that is useful for planning downstream.”

Hafner, for his part, isn’t convinced that more compressed representations of reality are better. “Yann is right about many things, ➤

8 August 2026 | New Scientist | 33

---

New Scientist audio

You can listen to many articles – look for the headphones icon in our app newscientist.com/app

even if maybe he expresses them more controversially than necessary, and the JEPA approach is very promising,” he says. “But I don’t think that representations should be compressed and tiny. Ultimately, you want to learn strong representations, and what we did with Dreamer 4 [which trains with pixels but predicts in abstract space] is incredibly robust.”

Intelligent thinking

More broadly, it is also far from clear at this stage if any of the world models in development – whether they learn and predict by reconstructing the world in high-fidelity or by inferring abstract representations – will ultimately be enough to get us to AGI.

Now, it’s fair to say that AGI is another elastic term, and that claims about world models as a route towards it exist on a spectrum, with some of the most ambitious suggesting that learned simulations of physical reality could become the foundation of generally capable reasoning agents. When it released Genie 3, for instance, DeepMind insisted world models are “a key stepping stone on the path to AGI, since they make it possible to train AI agents in an unlimited curriculum of rich simulation environments”. Indeed, Hafner argues that “if you have systems that are able to represent the rules governing the world, then you’re approaching something like AGI because that’s understanding, that’s a key part of intelligence”.

For his part, LeCun talks about them as a vital component of broader systems, rather than the whole story. Yet it is worth exploring the extent to which the kinds of world models in development would approximate human intelligence, because it can reveal what else might be required. Hafner says the most immediate requirement is temporal abstraction – that is, being able to reason not just about what will happen in the immediate future, but as things continue to play out. “You cannot just simulate everything at a sub-second, frame-by-frame level, because humans do not reason like that,” he says. “This is one of the open frontiers.”

Tenenbaum goes further. Human world models aren’t just engines for predictions, he says, “they are much richer than that”. Human reasoning depends on various forms of causal abstraction and hypothesis-driven inference –

“A limitation of AI systems today is a lack of awareness of their own cognitive state”

Our minds have models that let us predict what will happen next

not to mention models of other minds – which together allow us to flexibly recombine knowledge across different situations. “A key open question in all this is whether scaling up these world models will recover that richness, and my view is that it likely won’t,” says Tenenbaum.

Mitchell makes a similar, if slightly broader, point. “I think the ability to have a kind of compressed, simulatable representation of aspects of the world is very important, and this notion of world models is probably going to result in useful improvements,” she says. “But I think there are probably lots of other aspects of intelligence that matter.”

Another limitation of AI systems today, says Mitchell, is a lack of metacognition – an awareness of their own cognitive state, of what they know and don’t know, and how uncertain they are. “Is that fixable with a world model? Well, it depends what kind of world model, obviously,” she says. “But I’m a little worried that the notion of ‘world model’ is going to be used as the term for the difference between what we have now and ‘AGI’.”

All of which suggests that world models as currently conceived may well be necessary for human-level machine intelligence, but not necessarily sufficient. Indeed, while it looks increasingly likely that they will be powerful and genuinely useful in robotics, and possibly scientific simulation too, it is far from a sure bet that they will replicate the way world models work in human cognition, never mind human-level intelligence more broadly. “One of the big misconceptions is that intelligence is a single thing, that there is a single world model in the brain,” says Tenenbaum. “What we actually have is the ability to run many different models depending on the context, task and goal.”

Ultimately, then, the task of replicating what my 3-year-old daughter’s brain is capable of when it comes to modelling and predicting the physical world – even if it doesn’t stop her knocking glasses of water off the table – is not one to be underestimated. ■

Daniel Cossins is a journalist and former head of features at New Scientist

34 | New Scientist | 8 August 2026

---

New Scientist Discovery Tours

See the next total solar eclipse from the best vantage point on Earth

Science-led tours and cruises with astronomy experts, designed to put you on the line of totality

2027 Egypt | Morocco | Spain

2028 Australia | New Zealand

• Every journey meticulously planned for proximity to the eclipse's centre line
• Travel with leading astronomers and eclipse specialists who explain the science behind the phenomenon
• Enjoy the experience with other travellers who share your curiosity

Book early to secure your spot

Explore all eclipse tours and cruises newscientist.com/tours

---

Features

CSI: Microbiome

Every crime scene contains hidden microbial clues, but can we use them to solve cases, asks Matthew Ponsford

IN 2015, two intruders broke into Jose Lopez's home in Hollywood, Florida, and stole a TV. But this wasn't your standard-issue burglary. The "thieves" were deputies from the local sheriff's office, and Lopez, a biologist at Nova Southeastern University, was in on the whole thing as part of a project to see whether police can identify criminals from the microbes they leave at the scene.

Having analysed microbial traces left on surfaces and isolated bacterial DNA, the researchers could match them to the deputies. But when they tried to pick the intruders out of a database of 10,000 microbiome samples, the signatures proved insufficiently unique. The technique, they concluded in a 2020 paper, "cannot be used as a reliable trace evidence standard for criminal investigations".

Let's call that cause for reasonable doubt. Forensic scientists have long argued that the billions of bacteria, viruses and fungi in and on our bodies could become a valuable source of evidence. The problem has been proving that "microbiome forensics" is reliable enough for real investigations.

In recent years, however, researchers have refined their techniques, working with police to better establish what these methods can and can't tell us. They have also developed a way to analyse the post-mortem microbiome to estimate time of death – an approach that is being explored in relation to a controversial murder case in Italy.

"It's a massive frontier in forensic science right now," says David Carter at Chaminade University of Honolulu in Hawaii. The question is which of these techniques will prove useful – and whether the evidence they produce will ever stand up in court.

It's not difficult to see why forensic scientists see the human microbiome as an untapped source of information. "Imagine if I told you that there was a type of physical evidence that was guaranteed to be associated with every death scene – no matter where it was, when it was," says Carter, alluding to the many cases where fingerprints and DNA aren't retrievable. In those circumstances, the microbes we constantly shed onto other people and any

FOTOS: M. Ponsford

36 | New Scientist | 8 August 2026

---

surfaces we come into contact with could give clues because the ecosystems on our skin, for instance, vary hugely from person to person.

The field of forensic microbiology has really only existed since the 1990s, when advances in sequencing technology transformed what was possible in terms of analysing microbial DNA to isolate signature bacterial species. As sequencing got faster and cheaper, researchers could start testing the idea that the highly individual mixture of microbes we carry around – our microbial “fingerprints” – might be useful in identifying people.

For a study published in 2010, for example, Rob Knight, now at the University of California, San Diego, and his colleagues showed that the cocktail of bacteria deposited on a personal keyboard or mouse could identify the owner out of a database of 270 people. Other studies, meanwhile, showed that people leave a unique microbial signature in their homes.

This seemed promising. Yet when researchers tested the “touch microbiome” approach under more realistic conditions – including in those mock burglaries – it fell well short of the reliability required of forensic evidence.

The results prompted Jack Gilbert, then director of the Microbiome Center at the University of Chicago, who had led the burglary studies and others, to warn the US Department of Justice against adopting the technique in 2018. “I stipulated that I had no confidence that this could be used to identify individuals with enough rigour to definitively state that it was them and not somebody else,” says Gilbert. “I cannot prove, without a shadow of a doubt, that the bacteria that I found didn’t come from another human being.”

The setback didn’t end interest in forensic microbiology, though. Other researchers have since been exploring not only the extent to which better sequencing techniques and larger datasets can make a difference, but also whether other forms of microbiome analysis might play a role in criminal investigations.

Noemi Procopio, a forensic scientist at the University of Lancashire, UK, is investigating one such possibility. She argues that touch microbiome techniques might already hold

“investigative intelligence” that could give police crucial clues in cases where conventional forensic evidence is missing.

Her current research project began with the frustrations of Andy Nixon, a veteran crime-scene specialist for West Midlands Police in the UK, who is often confronted with murder cases where there is what he calls a “forensic void”. Many of the most vexing cases involve strangulation, says Nixon, because “the neck is too soft and elastic to preserve conventional fingerprints, and touch alone rarely leaves usable fragments of DNA – even when attackers use such extreme violence”.

Investigative toolkit

To see whether microbial traces might help, in 2025, Procopio and Nixon recruited 100 new officers from West Midlands Police to test how reliably their touch microbiomes could be traced back to them. The participants pressed their fingers onto acetate sheets, which were left around for days or weeks in a room where passers-by could walk through, as real-world microbial fingerprints might be. To start with, Procopio and her colleagues focused on rare bacterial species, which sometimes identified individuals but often became undetectable through sequencing within 48 hours, which is likely to be because they are overwhelmed by other species from the environment.

Next, the researchers analysed the overall structure of the microbial community in the samples to see whether it could offer a more reliable signature. “Maybe we have the same bacteria, but I have more of that, less of that, something unique,” says Procopio. In yet-to-be-published research, they found that this allowed them to match the sample to the individual with much greater accuracy, and that this signal remained a month later.

Crucially, Procopio isn’t claiming that such techniques will be a replacement for DNA analysis, sufficient alone to place a suspect at a crime scene. Instead, she sees them as “complementary” additions to the investigative toolkit. In fact, she agrees with Gilbert that “it will be very difficult, at least

8 August 2026 | New Scientist | 37

---

at this stage, to achieve the type of statistical weight typically associated with human DNA identification". But in cases where investigators draw a forensic blank, they might get a lead if the overall structure of a microbial community matches a sample taken from a known suspect.

Knight says this reflects how forensic evidence is often used in practice. The most common form of forensic finding is not one conclusive method of identification, but "class evidence" that narrows the suspect pool. Knight points to a 1982 case involving child murders in Atlanta, Georgia, where investigators found a mixture of microscopic fibres – yellowish-green nylon and violet acetate – on several murder victims and also in an unusual carpet owned by suspect Wayne Williams. They were able to explain to a jury the long odds of such a carpet being found in a suspect's home: 1 in 7792, based on sales records. "No individual fibre was conclusive, but the aggregate was very convincing," says Knight. "Treating microbes in this context is reasonable."

Microbial traces could have a similar role.

Fingerprints and DNA aren't always present at crime scenes

"Microbiome profiling could reveal clues about an unknown murderer"

"A technique doesn't have to be perfect to be useful," says Knight. "It's about stacking multiple class characteristics so that the probability of a false match approaches zero."

The microbial communities we carry around with us might also point in directions that investigators weren't looking. That's because our microbiomes are shaped to a great extent by our lifestyles and our interactions with others over the course of our lives, from where we live and who we spend time with to the foods we eat. In 2023, for instance, Procopio and her colleagues showed that microbiome

analysis can distinguish the residents of two neighbouring Italian regions, Lombardy and Piedmont, based on their oral microbiome. "This is definitely something that can be useful," says Procopio.

The challenge here has to do with reference data. For the moment, we don't have datasets of diverse individuals' microbiomes that are sufficiently large to draw useful conclusions. As with DNA, big databases make better insights possible. But Procopio reckons we could eventually use microbiome profiling to reveal clues about the lifestyle of an unknown murderer, for instance – whether they work in an office or on a farm, say, whether they have particular health problems or even whether they have pets.

Then again, for all the intrigue around efforts to use microbes to identify perpetrators from a crime scene, the most promising way to use the microbiome in forensics is in revealing when – and, in some cases, how – someone died.

Procopio is among a group of researchers seeking to use the bacteria that come and go as bodies decompose to develop microbial clocks that could offer a reliable estimate of the time since death, days or even weeks after a body is discovered. Few illustrate the potential better than one of the most controversial recent murder investigations in Italy: the case of Liliana Resinovich.

In January 2022, almost a month after the 63-year-old was reported missing, Resinovich's body was found on the grounds of a former psychiatric hospital in Trieste. She was found with plastic bags around her head and body, yet forensic pathologists found no obvious signs she had been attacked; their first report concluded that she had died, possibly by suicide, within the previous 48 hours.

But later, following protests by Resinovich's family, her body was re-examined. This new forensic assessment concluded not only that she had been killed by asphyxiation, owing to the identification of signs of trauma on the face and head, but also that Resinovich had been killed around the time of her disappearance – citing evidence including the undigested contents of her last known meal from the morning she vanished.

"Someone says she was dead for 48 hours, then the next forensic expert said she'd been dead for 23 days," says Procopio. "It's a massive gap." Crucially, it leaves open the question of what really happened: did Resinovich die two days before her body was found, as the first report suggested? Or was she killed around the time she went missing and kept somewhere cold, possibly even frozen, before her body was

38 | New Scientist | 8 August 2026

---

We all leave a unique bacterial signature in our environment, often from our fingers

moved to the place where she was discovered – a hypothesis proposed to explain why it appeared unusually well preserved for a body left outdoors for three weeks?

Despite what detective dramas might suggest, conventional forensic tools can pinpoint the time of death only when a body is discovered quickly. For the first 24 hours, a forensic pathologist can usually calculate a reliable “post-mortem interval” using a temperature reading from a rectal thermometer, says Procopio, while colouration, rigor mortis and cues from insects that lay eggs in the body’s orifices can help in the first few days. But uncertainty increases as the body enters more advanced states of decay, she adds, especially if it is outdoors.

This is precisely the kind of uncertainty that microbial clocks are designed to address, says Procopio. In the past decade, studies on human remains at “body farms” across the US have shown that decomposition proceeds in a particular way: different bacteria arrive, flourish and decline in predictable patterns. By tracking these microbial changes, researchers can estimate how long a body has been decomposing, potentially unearthing vital clues about when – and how – someone died. In the Resinovich case, a mouth swab taken

around the time of her body’s discovery preserved a snapshot of her oral microbiome. If her body had been frozen or kept somewhere cold before it was found, the microbial community in that sample should contain evidence of alterations consistent with a freeze-thaw cycle. “It would leave a detectable signature,” says Knight. Procopio adds that other clues might come from insects, because a study published in March showed that they introduce certain decomposing microbes to a body at different times depending on whether it was exposed outdoors from the moment of death or was moved, from inside to outside.

In Procopio’s view, the microbiome can resolve the uncertainty between the freezer hypothesis and the suicide hypothesis. All the evidence so far is “clearly going in one direction and not in the other”, she says. But for the evidence to be admissible in court, judges require compelling proof that microbial clocks are reliable. Which is why Procopio and her colleagues are now recreating the two scenarios at a body farm in northern Michigan, where climatic conditions resemble those in northern Italy. There, four donated cadavers are kept under different conditions, allowing the researchers to track how their microbiomes change over time with a view to presenting

a clear and admissible finding to the public prosecutor in charge of the Resinovich case.

On the broader question of whether microbiome forensics will ever produce evidence that leads to a criminal conviction, the answer depends on the application. In most countries, judges act as gatekeepers, with the power to decide if a new form of scientific evidence should be allowed in courts, often following guidelines similar to those set down in the US starting in 1993. Under the Daubert standard, as it became known, judges consider whether a technique has been independently tested and peer-reviewed and has a known error rate before admitting it as evidence.

On all counts, microbial clocks to estimate time of death appear ready to be used in court, says Carter. Research led by Jessica Metcalf at Colorado State University shows that a reliable post-mortem interval judgement can be made with an error rate of three days if a body is discovered in 10 days, though that gets larger the longer someone has been dead. “I feel that it meets those criteria, but it would really be a judge that would determine that,” says Carter.

The prospects for using microbiome analysis to identify suspects are more uncertain. To offer an alternative to DNA matching in the identification of suspects, it would have to achieve similarly exacting evidentiary standards – roughly a million-to-1 chance of a false match. The difficulty here is that the skin microbiome remains such a dynamic and ever-changing ecosystem that Gilbert doubts it will be useful any time soon. “The microbiome contains 500 species, maybe 5000 to 6000 strains that are constantly fluctuating in abundance, constantly undergoing genetic evolution, constantly shifting in the proportional representation and constantly being exchanged between people,” says Gilbert. “You’re dealing with biology, and biology is horribly complex.”

Others are more optimistic. Knight argues that sequencing technology has moved on in critical ways since 2018. More advanced methods can detect less obvious genetic differences between microbes, he says, revealing “personalised and persistent strains” that older approaches could not. “I wouldn’t give up on the trace evidence topic,” says Knight. “It’s challenging, but technology is advancing rapidly.”

Matthew Ponsford is a journalist based in London

8 August 2026 | New Scientist

39

---

Features Interview

“A grief attack is, essentially, a full-human-experience attack”

A few years ago, psychologist Sherman Lee experienced a “grief attack” and set out to study them. He tells Vanessa Nirode just how powerful and all-consuming they can be

40 | New Scientist | 8 August 2026

---

ONCE, while driving his car, Sherman Lee started sobbing so violently that he lost control of his muscles and had to pull over for fear of crashing. His whole body was subsumed in grief over a beloved pet who had died a month previously. He'd had panic attacks before, but this seemed different; he wasn't just breathless, but filled with intense despair.

Looking back now, Lee says the episode felt like a grief attack, a term that was coined decades ago and is widely recognised within psychology, but hadn't been rigorously researched. Last year, Lee, a psychologist at Christopher Newport University in Virginia and an expert on grief, co-published a study of almost 250 people who had experienced grief attacks to better understand how to define them.

Lee's research fits into a growing understanding of grief as a complex and nonlinear experience. The "five stages of grief" concept has long been debunked, and psychologists now emphasise that grief can last for years, affect both body and mind, and permeate every aspect of life. It is both intensely distressing and one of the most profound and intrinsically human experiences.

Speaking to New Scientist, Lee explains what he has uncovered about grief attacks, and the importance of understanding the myriad expressions of grief.

Vanessa Nirode: What are the different ways grief can manifest?

Sherman Lee: Name any facet of a person's life and I can show you where grief will express itself. Biologically, grief is shown in sleep behaviour or eating disturbances. A grieving person can have lower immunity because of stress. If a couple's been together for a very long period of time, there's a good chance the widow or the widower will follow soon after.

Psychologically, grief can produce strong emotions like yearning, sadness or anger. People become preoccupied with the person they lost, especially if it was a tragic death. A lot of people have hallucinations. In one large study of more than 10,000 people who've experienced grief, 50 per cent of participants

reported a sensation of a presence; one-third perceived hearing their voice.

Socially, people may avoid anything that triggers their memory of the person. Their lives get very small. They may avoid activities they used to do with the deceased. Or they may turn to coping through some form of substance abuse.

Spiritually, you can be angry at God, not believe in the things you used to believe in – a very big existential crisis.

This concept of a grief attack may not be familiar to many of us, but I believe it has been around for a while.

Therese Rando is a pioneer in grief therapy. She first explored the subject of grief attacks in 1991 [though she called them "sudden upsurges of grief" at the time] and has written extensively about it based on clinical observations. In contrast, my colleagues and I identified and scientifically validated the symptoms using data drawn from large numbers of people.

So, what exactly is a grief attack?

We were able to identify four main characteristics: panic (which manifests with dizziness, shortness of breath and fear), yearning, despair and incoherent thinking. We also created a tool to recognise and evaluate the symptoms. [Grief attacks] are different from "standard" grief, as they include physical symptoms as well as emotional ones and at an intensity that usually comes on suddenly. We don't yet know how common they are in terms of frequency, but they are mentioned, often by different names such as "pangs" or "spasms", in many books and articles that I've read. I think all this gives us an indicator that it's relatively common.

The most significant thing is that this is a real condition, and as such can be scientifically studied and, if necessary, treated.

How are grief attacks different from panic attacks, and why make the distinction?

From a basic science perspective, it brings together two systems that are often studied separately: the fear system and the attachment system. Grief attacks look like a unique storm of these two systems occurring at around

8 August 2026 | New Scientist | 41

---

the same time, with sweating and shaking as manifestations of the fear system, and yearning thoughts as part of the attachment system.

When you have a grief attack, along with typical panic attack symptoms like shortness of breath and dizziness, you also experience acute grief symptoms such as yearning, irresistible crying and thoughts about the deceased. The intensity of these two things together makes it a unique class of human experience.

Do we know why some people experience them and others don't?

I would postulate that two main things might contribute.

Firstly, a person [might experience grief attacks if they have] anxiety sensitivity, which is the fear that anxiety symptoms like a racing heart are physically harmful. For instance, if your heart races or you can't breathe, some people interpret that as a threat to their life, so they get really scared and panicky. I happen to fall in this spectrum. We know that people who have anxiety sensitivity are five times more likely to be diagnosed with a panic disorder and, because grief attacks contain a panic attack component, it would follow that they are also more likely to experience grief attacks.

The second factor is stress build-up, which can trigger panic attacks and make grief symptoms worse. If a person has compounding stress in their life, they are probably more likely to have a grief attack.

How long are people deeply affected by grief and grief attacks?

I think when we say "deeply affected", what most people mean is how long grief is dysfunctional or extremely distressing. Grief researchers give it about six months. The latest Diagnostic and Statistical Manual of Mental Disorders (DSM) gives a whole year. After that, we start talking about prolonged grief disorder.

The way I look at it, though, "deeply affected" means changing you in a fundamental way. I think it's lifelong. Take C.S. Lewis. He wrote a whole book about his experience of losing his wife. Many people can function, laugh and have fulfilling lives so that when you talk to them, you wouldn't know that they lost something significant, but every time they paint or write, it's there.

[Grief attacks] can occur anywhere along a course of one's life post-loss, though they

"Those who allow the grief attack to come and go fare better than those trying to distract themselves"

seem to [be] more frequent early on in the grief process.

Is there anything people can do to manage the experience of a grief attack?

We asked people how they coped, though we didn't directly do any experiments. Our research suggests that those who allow the experience to come and go fare better than those trying to distract themselves or focus on some activity. It's what mindfulness teaches us: don't judge it, don't fight it, just allow the experience to come because it will leave in a few minutes.

What do we know about how grief affects the brain and what happens in our brains during a grief attack?

Cognitive tests show that the more complicated and intense the grief experience is, the more impairment and disruption we see in memory and attention. A few studies show smaller brain volume in people with very severe grief. Also, chronic stress leads to neuronal loss, which could be reflected in brain volume. It also hits the brain area that deals with rewards and addictions, as well as regions associated with physical pain.

42 | New Scientist | 8 August 2026

---

The loss of a loved one can lead to sudden panic-attack-like episodes

MARCUS BATTEL/MILLENNIUM IMAGES, UK

Grief attacks can include feelings of intense yearning and despair

During a grief attack, multiple areas of the brain activate at the same time. Because of the fear aspect, the amygdala is involved, similar to a fight-or-flight response. The periaqueductal grey [an area of the brain that regulates pain modulation and flight-or-freeze responses] kicks in because you yearn for that person to be back. The brain reward system's nucleus accumbens, which searches for the person in a craving sort of way, gets triggered. And then there's the anterior cingulate cortex, the pain centre of the brain, where the experience of pain and emotion come together. A grief attack is, essentially, a full-human-experience attack.

How can you know if grief or grief attacks are becoming worrying, as opposed to a natural process?

I think what we mean by "natural" is good: what is good or healthy grieving? Which, in a weird way, assumes that good has to be pain-free. You love someone and the natural consequence of that love is the pain that comes from disconnection. You have to deal with that pain until it subsides.

While grief attacks are disruptive enough to be scary, I wouldn't say they're worrisome

or directly physically dangerous. After mine, I was scared enough that I didn't want to drive [for] the rest of the day, but I did have enough control to pull over. I think most people [in our study] were the same way. Psychologically, they can feel like you're losing control, which is scary, but not necessarily dangerous. At the same time, mental events are disruptive and real. We shouldn't ignore them and people should ask for support if they need it.

You should seek a professional whenever you find that grief is too much to handle, when it impairs your ability to function.

It sounds like relatively few of us will experience a grief attack, but almost all of us will feel grief at some point in our lives. What is your advice for people in that situation?

For a lot of people, closure is important: rituals, a funeral, being able to have final words. Talking to people about the loss is helpful when trying to make sense and gain perspective. There is evidence that shows that physical movement, exercise and sport are helpful for those mourning a loss. As well as combating loneliness and depression, exercise and sport can help with sleep problems, especially if you get exhausted

enough to override your overthinking mind. Also, if you're doing a sport, you're not at home looking at pictures; you're getting a temporary distraction. We all need a break from pain.

When you decide to seek out a professional, grief therapy is tailored for you individually. Because grief is a big thing. It's never one thing. Many people might suggest cognitive behavioural therapy, which deals with the thoughts and actions that give you problems. Let's say what bothers you are the grief attacks. Then I would suggest acceptance-based therapies, which aim to teach you to let them happen and accept them, thus building psychological flexibility and a way to keep yourself grounded in the present.

All of these are great therapies and approaches, but it's hard to give a silver bullet for [grief]. It has many parts. I suggest trying to figure out the problems that you are having within that big thing. ■

Vanessa Nirode is a journalist who writes about health and wellness

8 August 2026 | New Scientist | 43

---

THE DIFFERENCE IS NEVILLE JOHNSON

Transform your existing staircase with no mess, no fuss, and best of all, no building work.

Renovating your staircase into an exquisite feature is much easier and less costly than you'd imagine. In just 1-2 days and with no building work, we can transform the heart of your home with a stunning, handcrafted staircase. From glass designs that make spaces feel lighter and more open, to elegant steel details that elevate the overall look of your staircase, every design is tailored to suit your home, style and budget.

✓ The UK's market leader in staircase renovations ✓ Installed in less than 48 hours by experts ✓ Made to order with quality materials ✓ 10 year guarantee ✓ Finance options available

REQUEST A BROCHURE 0161 873 8333 | NEVILLEJOHNSON.CO.UK

---

The back pages Puzzles

Quick crossword #214 Set by Richard Smyth

Scribble zone

Cryptic crossword next week

ACROSS

1 Automatons (6) 4 Medical imaging procedure (1,1,4) 9 Notion, invention (4) 10 Amphetamine trade name (10) 11 Lymphoid organ in the throat (6) 12 Lynn ___, US evolutionary biologist (8) 13 Cluster of boils (9) 16 Skin condition that causes pimples (4) 17 Circle (4) 18 1 million millimetres (9) 21 Scuba brand co-developed by Jacques Cousteau (4-4) 22 Ag (6) 25 Geological era that began around 540 million years ago (10) 26 Accessible (4) 27 Phenomena of flowing water (6) 28 Reciprocal function, in trigonometry (6)

DOWN

1 Broadcast technology associated with Guglielmo Marconi (5) 2 Alloy of Cu and Zn (5) 3 Cell protein (7) 5 Supposition or hypothesis (6) 6 Blood cell, perhaps (9) 7 Of a system, having outputs that are disproportionate to inputs (9) 8 Taxonomic grouping to which Homo sapiens belongs (6,7) 14 Even (3,6) 15 A virus, toxin or bacterium, for example? (9) 19 Small bone of the middle ear (7) 20 Wrinkled (6) 23 Dopamine precursor (1-4) 24 Discharge from the body (5)

Quick quiz #364

set by Chelsea Whyte

1 Hong Wang became the third woman to ever win the Fields medal, the most prestigious prize in mathematics. What did she win it for solving? 2 Which ingredient of the cosmos was recently found to perhaps not exist? 3 Which marine animal were orcas seen ramming, causing them to explode into thousands of fragments? 4 An agricultural start-up developed a way to increase yields of soya beans, peppers and cannabis. What did it use to stimulate plant growth? 5 What discovery in Cloggs cave in Australia suggests the GunaiKumai people have been conducting elaborate rituals there for more than 1000 generations?

Block Universe

created by Puzzlr

The goal is to divide the grid into non-overlapping rectangles. Each rectangle must contain exactly one number, and have an area equal to that number. For example, a 6 should have a rectangle covering six squares.

Answers to this week's puzzles on page 47

Our games are now playable online newscientist.com/games

8 August 2026 | New Scientist | 45

---

The back pages Almost the last word

Branching out

My family tree doubles at every past ancestor generation, branching back into the past. At what point does it decrease to reflect the size of the smaller historical population?

Hillary Shaw

Newport, Shropshire, UK Global population was around 2 billion in 1930, 1 billion in 1820, 0.5 billion in 1640 and 0.25 billion in 900. Meanwhile, if your parents were born in 1950, and 10 doublings equal 1000 ancestors, theoretically you would have 2000 ancestors from 1700 (assume average maternal age 25), 2 million from 1450 and 2 billion from 1200. So, these two series cross around the year 1300.

However, your family tree numbers stabilise much later and smaller than that because, before railways, many people lived in isolated communities and married distant relatives.

As an extreme example, the Scottish archipelago of St Kilda had around 100 people for several centuries before mass emigration kicked in around 1800 – about the minimum number to avoid excessive inbreeding and genetic problems.

On the British mainland, before railways, there were travelling

“In the 1700s, chances are that most of your ancestors came from a village of a few 100. Before that, it’s almost all inbreeding”

merchants, justices, preachers, soldiers and beggars, but people would marry in a similar social class, religion, et cetera.

In the cities, especially port cities, you might marry an outsider, but pre-railway villages were almost as isolated as St Kilda and of similar population size.

Around 1650, it took seven days to travel between York and London, due to atrocious roads, which would have been quagmires

ARTESIA PETT LEE / BRIAN PHILLIPS

This week’s new questions

Fish-eye view Humans can see water but not air. So can fish see air? Kayla Zelenietz-Mills, age 13, Ottawa, Canada

Feeling negative Would it be possible to have anti-black holes, and how could we tell if we were looking at one? David Oxnam, Toppesfield, Essex, UK

in winter, dust bowls in summer.

In 1700, just 15 per cent of the British population lived in urban areas, which were small. London then had 650,000 inhabitants, but other cities were much smaller: Bristol had 20,000, Manchester 10,000 and Liverpool 5000 people. So, chances are that most of your ancestors came from a village of a few hundred in the 1700s. Before that, it’s almost all inbreeding.

Simon McLeish

Lechlade, Gloucestershire, UK It is actually more to do with the availability of unrelated partners than the wider population size. The availability of the bicycle has been cited as one factor that made it easier to find less closely related prospective partners. Parish records in Dorset, UK, show that,

between 1887 and 1916, the proportion of marriages between people in the same village dropped from 77 to 41 per cent.

The royal families of Europe often married close relations for centuries to preserve the purity of the royal blood. The Habsburgs were the best known example, with the “Habsburg lip” a physical trait seen in many portraits of those related to the dynasty.

Robin Maguire

Hobart, Tasmania, Australia The answer will vary when factors such as geography or social practices limit potential marriage partners to a particular segment of the population.

For an extreme example, consider the Ptolemaic dynasty of Egyptian rulers. After the death

Do fish, like this largescale four-eyes, see air, just as people see water?

of Alexander the Great in 323 BC, the Macedonian general Ptolemy I Soter appropriated the throne of Egypt and his descendants kept power in the family through the practice of intermarriage.

The final ruler in the Ptolemaic line was Queen Cleopatra VII (born in 69 BC), the paramour of Julius Caesar and subsequently of Mark Antony. Going back through her family tree, she had about the usual number of direct forebears in the late 100s BC, but all of these were exclusively descended from the irreducible minimum of just two ancestors in the late 200s, King Ptolemy V Epiphanes and his consort, Queen Cleopatra I. There were six incestuous liaisons along the way. This unusual background did not seem to do the final Cleopatra any harm: she was almost as famous for her learning as for her vivacity.

John Davies

Haverbreaks, Lancaster, UK Your family tree’s size just stops doubling when it reaches the size of the earlier population, and you discover that two of your great-to-the-X-grandparents are the same individual!

Anthony Woodward

Portland, Oregon, US It never does. Tracing your ancestors backward in time for 8 million years, approximately 32,000 hominin generations, they will become the last common ancestors of chimpanzees and hominins.

Energy savings

How does energy use by the body differ when I walk, jog or run for 1 kilometre?

Eric Kvaalen

Les Essarts-le-Roi, France An article in Nature gives a graph showing oxygen consumption as a function of speed for walking

Want to send us a question or answer?

Email us at lastword@newscientist.com Questions should be about everyday science phenomena Full terms and conditions at newscientist.com/lw-terms

46 | New Scientist | 8 August 2026

---

Tom Gauld

for New Scientist

and running ("Endurance running and the evolution of Homo", 2004).

For walking, the curve is U-shaped, with a minimum of about 165 millilitre O₂ per kilogram body weight per kilometre at around 1.3 metres per second. Running (which includes jogging) is a straight line that decreases from around 210 ml O₂/kg/km at 2 m/s to around 200 at 6 m/s.

This means that if you want to go a certain distance in a given, fairly long amount of time while consuming the lowest possible amount of oxygen, then you should walk.

However, when the allotted time is less than a certain value, corresponding to a slightly fast walk, then you should alternate between running as fast as you can and walking at that slightly fast walking speed.

Note, however, that this is talking about total oxygen consumption, not extra oxygen consumption over normal basal rate. To minimise extra oxygen

"My theory is that, because dogs don't climb and have no aerial predators, they tend to not have much concept of 'up'"

consumption, the optimum is a slower walk than that for total consumption.

Hear that?

Why don't dogs look up when a low-flying plane or helicopter passes overhead? I've never seen one do this. (continued)

Dyane Silvester

Arnside, Cumbria, UK As a former dog owner, and now regular dog-sitter, my theory is that, because dogs don't climb and, as a species, have no aerial predators, they tend not to have much concept of "up" (except where it's upward-sloping ground).

If you drop something from a height, most dogs will look for it

on the floor rather than realise that there was a table between the object and the floor.

I did, however, have one Alsatian who, if I appeared at an upstairs window when she was in the garden, could immediately locate and arrive at the room I was in, despite having never been able to see out of any upstairs windows in our house. She also noticed and watched birds and low-flying aircraft.

I did manage to stump her once: by climbing out of the skylight and appearing on the roof. She never worked that one out!

David Lewis

London, UK

Although unusual, a friend's late spaniel would look at aircraft flying above her in the park. She would then bark and chase them. Fortunately, she never caught one!

C. H. Johnson

Collyton, Devon, UK

Because they know planes and helicopters don't bring their food or any doggy treats. ■

Answers

DOWN 1 Radio, 2 Brass, 3 Tubulin, 5 Theory, 6 Corpuscle, 7 Nonlinear, 8 Animal kingdom, 14 All square, 15 Biohazard, 19 Ossicle, 20 Rugose, 23 L-dopa, 24 Egest

27 Eddies, 28 Secant

25 Palaeozoic, 26 Open,

21 Aqua-Lung, 22 Silver,

16 Acne, 17 Loop, 18 Kilometre,

12 Margulis, 13 Carbuncle,

9 Idea, 10 Benzedrine, 11 Tonsil,

ACROSS 1 Robots, 4 CT scan,

#214

Quick crossword

5 Ash from burnt grass

4 Sound waves

3 Sunfish

2 Dark energy

1 The Kakeya conjecture

Quick quiz #364

Solution

Block Universe

8 August 2026 | New Scientist | 47

---

The back pages Feedback

Bloons and doubloons

This edition of Feedback is an exercise in time travel – in that every item is something from absolutely ages ago that we somehow missed. We would feel guilty, but it is the summer holidays (in the northern hemisphere, anyway) and many readers will have half an eye on their errant children, so we thought we could get away with it.

So, cast your minds back to 2008. This, obviously, was the year of the 20th Canadian Conference on Computational Geometry, held in Montreal. A large amount of research was presented at this conference, but reporter Matthew Sparkes recently stumbled across a theoretical study by Erik Demaine, Martin Demaine and Vi Hart. It presents “a general mathematical and algorithmic theory for balloon-twisting structures”.

If you didn’t grasp that, allow us to clarify: it’s about the geometry of balloon animals. The authors develop a mathematical framework that allows them to calculate how many balloons you need to generate a particular shape, whether it is a dog (one) or some complicated polyhedron like a dodecahedron (10). These, naturally, are referred to as “Platonic balloons”.

Throughout the paper, the authors use the term “bloon” to refer to a mathematical entity equivalent to a balloon. Feedback wasn’t clear why this neologism was necessary, until we realised that they were therefore able to refer to a balloon of twice the usual length as a “doubloon”.

Curious to know what the authors had been up to since, Feedback went a-searching. Hart became a science populariser on YouTube, then a senior manager at Microsoft. Meanwhile, Martin and Erik are father-and-son researchers and artists, both fascinated by the mathematics of folding.

Erik, a professor at the Massachusetts Institute of Technology, has been in New Scientist a lot. Recent papers include “Overlapping unfoldings of cones and convex polyhedra” and “Tetris

Twisteddoodles for New Scientist

Got a story for Feedback?

Send it to feedback@newscientist.com or New Scientist, 9 Derry Street, London, W8 5HY Consideration of items sent in the post will be delayed

is hard with just one piece type”. More from Erik, please.

Testing times

Continuing on the topic of being very late to things, reader Derek Long flagged a paper from 2018, “Parachute use to prevent death and major trauma when jumping from aircraft: randomized controlled trial”. It reports on a study called “Participation in Randomized trials Compromised by widely Held beliefs about lack of Treatment Equipoise (PARACHUTE)”.

(Side note: that is a truly excellent scientific acronym. We haven’t had any of those in a while.)

The problem posed is one of evidence. “Parachutes are routinely used to prevent death or major traumatic injury among individuals jumping from aircraft.

However, evidence supporting the efficacy of parachutes is weak and guideline recommendations for their use are principally based on biological plausibility and expert opinion.” If readers haven’t yet twigged: the paper is satirical, poking fun at the idea that even the most common-sense interventions should be tested in a full randomised controlled trial.

As the authors write: “In seminal work published in the BMJ in 2003, a systematic search by Smith and Pell for randomized clinical trials evaluating the efficacy of parachutes during gravitational challenge yielded no published studies.”

Feedback actually ran across that review back in 2004. The authors noted they were “unable to identify any randomised controlled trials of parachute intervention” and suggested

that “everyone might benefit if the most radical protagonists of evidence based medicine organised and participated in a double blind, randomised, placebo controlled, crossover trial of the parachute”. Oof.

Fifteen years later, they did do a trial. This showed that “parachute use did not significantly reduce death or major injury”, with the slight caveat that “the trial was only able to enroll participants on small stationary aircraft on the ground”. This was helpfully illustrated by a photo of one of the participants jumping, parachute-less, out of a grounded plane. As the authors say: “Our findings should give momentary pause to experts who advocate for routine use of parachutes for jumps from aircraft in recreational or military settings.”

Witty Zwicky

Some concepts can’t be traced back to their origins. One such is the spherical cow joke, which illustrates the difficulties of working with highly simplified and abstract models of complicated, real-world situations. It involves a physicist who figures out how to improve productivity on a dairy farm, but their solution only works for spherical cows in a vacuum. Nobody seems to know who came up with this: the oldest references place it in the 1960s, but the origin stories are all vague and one reference replaces the cows with chickens.

However, another sphere-based physicist joke seems to be older. Feedback has just learned, many decades after the fact, about an insult deployed by Fritz Zwicky, who, in the 1930s, produced some of the first evidence for dark matter. A noted curmudgeon, Zwicky was fond of referring to people he disliked as “spherical bastards”. Such a person is a bastard no matter which way you look at them.

Feedback will be using that from now on. But we also want to know the etymology (if that is quite the term) of spherical bastards, chickens and cows. Any insights welcome. ■

48 | New Scientist | 8 August 2026

---

Summer

entertaining made effortless

Save £401

when you purchase any Quooker system with a CUBE and installation by a Quooker engineer

Offer ends 31st August 2026

Discover how Quooker can transform your Summer.

Enjoy instant 100°C boiling, chilled, and sparkling water – all from one tap. A Quooker tap can be installed in your existing kitchen in under two hours*, is simple to use, and impossible to live without.

Time-saving. Energy-efficient. Safe.

Quooker transforms everyday kitchen tasks while it can help you save time, reduce waste, and help to improve safety.

See it in action – from your own home.

Book a free live virtual demonstration with our expert team. They'll walk you through the features and answer any questions, one-to-one.

Scan the QR code to unlock exclusive offers and to arrange a virtual demonstration or visit quooker.co.uk

Experience Centre

Beaumont Buildings, Great Ducie Street, Manchester, M3 1PQ

“

Quooker isn't just a tap, it's a game changer in the kitchen, and the kind of company that makes you feel good about where you've spent your money.

”

Quooker®

100°C BOILING, CHILLED AND SPARKLING WATER

The tap that does it all

*Installation time may vary depending on kitchen setup.

---

Speedmaster

Iconic since 1957

SPEEDMASTER GREY SIDE OF THE MOON

Co-Axial Master Chronometer

Ω

OMEGA

omegawatches.com

QA Report - New Scientist - 08.08.2026.pdf

Pipeline

• Extraction route: mistral-ocr
• extraction_route=mistral-ocr
• Translation chunks: 69
• Extraction warnings:
• [OCR-ROUTE] FORCE-MISTRAL: All PDFs route directly to MistralOCR.

Non-model Checks

• WARNING: missing numeric tokens from source: 10003, 10800, 1864, 1887, 20000, 230, 26000, 3066, 31712, 323
• WARNING: Missing proper nouns in translation: Alexandra Thompson, Alison Flood Senior, Alison George, Amandeep Sian, Amazonian Research, Analysis Environment

Is, Ancient Mariner, Anthony Woodward

Portland, Aperture

Striking, Archaeology

Smallpox - WARNING: source dates, money, or percentage tokens may be missing

Model QA

经过严格审核，该译文存在多处致命性数字错误（量级错误）以及严重的内容错位/幻觉。以下是详细审计报告：

═══ 审核报告 ═══

1. 数字量级错误 (BLOCKING)

• 捐款金额错误：
  • 原文：Will you donate £10
  • 译文：您愿意捐赠£1万
  • 结论：[BLOCKING] 严重量级错误，将 10 英镑误译为 1 万英镑。
• 百分比量级错误：
  • 原文：about 90% used for firefighting, and only 10% for prevention
  • 译文：约 90% 用于灭火，而只有 1万% 用于预防
  • 结论：[BLOCKING] 严重量级错误，将 10% 误译为 10,000%。
• 页码/索引数字错误：
  • 原文：10 Quantum advantages
  • 译文：1万 量子优势
  • 结论：[BLOCKING] 将页码 10 误译为 1 万。

2. 内容错位与幻觉 (BLOCKING)

• 段落内容完全错位（严重）：
  • 译文在“家谱”讨论段落中突然出现了：“在 1700 年代，很有可能你的大多数祖先都来自一个只有几 100. 人的村庄。在那之前，几乎全部是近亲繁殖”商人、法官、传教士...
  • 译文在“步行/跑步耗氧量”段落中突然插入：“我的理论是，因为狗不会攀爬且没有空中捕食者，它们往往没有太强的‘上方’概念”
  • 核查：经核查“抽样原文”，这两段话分别属于不同的文章（一个是关于家谱的 $\text{Almost the last word}$，一个是关于狗行为的 $\text{Feedback}$ 问答）。译文将两篇完全无关的文章内容在句子中间进行了物理拼接/交叉污染。
  • 结论：[BLOCKING] 严重的翻译逻辑崩溃，导致正文内容完全不可读。

3. 完整性与残留问题 (WARNING)

• 低中文密度区域未翻译：
  • 译文末尾的“答案”部分（纵向/横向填字游戏答案）仅进行了简单的罗列，未对部分术语进行翻译或标注，且格式混乱。
  • 结论：[WARNING] 虽属于过滤版面数据，但作为交付物，其拼接方式过于粗糙。
• 术语翻译不统一：
  • agency 在文中被翻译为“能动性”，但在部分段落中直接保留英文或翻译不自然。
  • 结论：[WARNING] 影响阅读流畅度。

4. OCR 根源问题 (SOURCE_OCR)

• 数字符号干扰：
  • 译文中出现大量 1.08万,800、1万,000 等奇怪的数字格式。
  • 核查：原文为 10,800 和 10,000。这显然是 OCR 将逗号识别为某种分隔符，而翻译软件在处理时错误地将其转换成了中文的“万”字。
  • 结论：[SOURCE_OCR] 根源在 OCR 识别与自动转换逻辑。

═══ 最终审计结论 ═══

状态：[FAILED] 主要问题： 译文出现了极其严重的数字量级错误（£10 $\rightarrow$ £1万）和内容交叉污染（将两篇不同文章的句子强行拼接在一起）。该版本完全不可用，必须重新翻译。