2026-02-13 21:30:00
2007年，美国俄勒冈健康与科学大学（OHSU）的灵长类动物研究中心被PETA的调查员拍摄到猴子的图片。| PETA 特朗普政府的科学议程被广泛认为是对科学进步的战争，但请听我说：事情还有更多层面。该政府的科学政策并非仅仅由反科学意识形态主导，而是由一群具有不同批评意见的多元势力共同塑造，他们愿意放弃既有的科学传统。其中包括一些动物权益倡导者，他们本身也是科学家，合理地希望推动科学超越目前对动物实验的依赖。
目前，从老鼠、兔子到猴子等动物仍支撑着大量医学研究。但它们作为人类研究模型的效用一直有限。正如哈佛生物工程学家唐·英格伯去年所说：“大家都承认，动物模型在最好的情况下也只是一般，更常见的是高度不准确。”对动物实验的伦理问题也极为严重，同时，一种新的无动物研究技术正在迅速发展，包括实验室制造的类器官、器官芯片和先进的计算建模。
基于这一思路，美国国家卫生研究院（NIH）——美国大学生物医学研究的主要资助者——在去年由主任杰伊·巴特查里亚领导下宣布，计划优先采用无动物实验方法，并减少其资助研究中对动物的使用。同时，NIH与一所主要的美国生物医学研究大学合作，迈出了实现这一目标的重要一步。本周，OHSU的董事会一致同意开始与NIH就其提议结束灵长类动物实验、将研究中心转变为动物庇护所进行谈判。许多反对动物实验的人希望此举能推动对灵长类动物实验的逐步淘汰。
OHSU的灵长类动物研究中心是美国七所联邦资助的灵长类动物研究中心之一，目前饲养着约5000只猴子，包括恒河猴、日本猕猴、狒狒和松鼠猴等，占美国所有灵长类研究猴子的约5%。根据本周达成的决议，该中心将停止繁殖新猴子，除非现有实验需要，同时在未来六个月与NIH讨论可能的转型计划，将研究中心从灵长类动物繁殖和实验设施转变为庇护所。
OHSU长期以来因动物福利问题备受争议，过去几十年中曾多次被联邦动物福利法违规。例如，2020年一名工作人员不小心将两只猴子放入笼子清洗机中，导致它们死亡；2023年，一只新生猴子被掉落的滑动门砸死。维琴尼亚法律与研究生院动物法律与政策研究所的教授德尔西安娜·温德告诉我：“OHSU的记录是我见过最差的，他们总是不断发生疏忽导致的死亡。”（披露：2022年，我曾参加过该学院的一个媒体交流项目。）
本周一，OHSU灵长类研究中心的研究人员以及来自大学和公众的其他人士在公开会议上激烈辩论了结束该中心研究的提议。俄勒冈州急诊科医生迈克尔·梅茨勒表示：“尽管过去灵长类动物研究可能对医学进步有所贡献，但如今先进的方法已使其几乎过时。”他指出，这些猴子研究分散了资金和注意力，无法与以人类为中心的研究相比。
支持该研究中心的人则批评大学“在政治压力下立即向敌对政府屈服”，如OHSU的一名生物医学工程博士生科尔·贝克在听证会上所说。OHSU显然面临来自NIH的压力，因为NIH自2023财年起提供了该大学大部分研究资金。白宫也表明，它愿意惩罚不配合其政策的大学。然而，呼吁关闭该中心的呼声早在特朗普政府之前就已存在，而且并不只是共和党的优先事项。俄勒冈州民主党州长蒂娜·科特克曾呼吁关闭该中心，她以哈佛大学为例，哈佛大学在2015年因对猴子的待遇问题而关闭了自己的灵长类动物研究中心。哈佛的决定本身就是一个重要的信号，表明医学研究正在向新的方向发展。
为什么我们仍然在灵长类动物身上进行实验？关于灵长类动物研究必要性的争论往往难以理解。支持者和反对者似乎在使用不同的语言。反对者认为动物数据对人类应用价值有限，而支持者则坚持认为没有猴子就无法研究人类的严重疾病。20世纪的科学史学家托马斯·库恩曾用“不可通约性”一词描述这种沟通上的断裂。在不同范式下工作的科学家即使面对相同的现象，也可能得出截然不同的结论，因为他们从不同的概念框架来看待问题。正如前OHSU教授、Vox撰稿人加雷特·拉希维斯所指出的，科学家们往往被孤立，这种专业化的研究方式使得他们难以从更广泛的科学视角看待问题。
灵长类动物研究，像科学中的许多领域一样，是路径依赖和历史环境的产物。20世纪60年代，美国建立了联邦资助的灵长类动物研究中心，如OHSU的中心。当时NIH认为灵长类动物实验是未来的方向，这一观点至今仍影响着医学研究的实践。然而，如今实验室中笼养的猴子更像是一种过时的象征。显然，至少有些灵长类动物实验在美国家实验室中的作用极为有限，特别是那些试图通过在猴子身上诱导抑郁等复杂心理疾病来模拟人类情况的研究。
2014年，前NIH主任弗朗西斯·柯林斯在一封被PETA获取的私人电子邮件中承认了这一点，称“在非人类灵长类动物身上进行的大量研究毫无意义”。此外，灵长类动物的圈养状态可能使实验结果更难以应用于人类。例如，拉希维斯认为，极端的笼养环境会损害实验动物的健康，并扭曲猴子的心理状态，使它们难以作为健康人类的可靠替代品。
尽管支持灵长类动物研究的人引用其在人类药物开发中的应用，如艾滋病疗法，但灵长类动物数据的存在并不能证明这些研究是不可或缺的。考虑到对社会和认知复杂的动物进行研究的高道德风险，以及将资源和职业生涯投入到灵长类动物实验室的高昂机会成本，仅仅偶尔有用似乎不足以证明将猴子置于终身圈养和侵入性实验中的正当性。
NIH值得肯定的是，它正在采取这一观点。事实上，过去十年中，联邦政府已经停止了对黑猩猩的生物医学研究，尽管其他灵长类动物在研究中的参与程度比黑猩猩更深。因此，NIH现在面临的挑战是如何在尊重研究人员职业发展的前提下逐步减少灵长类动物研究；建立一个可信的无动物研究工具的过渡方案；并在其资助灵长类动物庇护所的提议中，为在联邦资助科学中受到伤害的动物提供一定程度的正义。这对于任何正常政府来说都是一个艰巨的任务，而对于一个已严重损害科学界信任的政府而言，更是难上加难。这可以被视为一个测试案例，看看特朗普政府在削减科研经费的同时，是否能够推动至少一次积极的科学范式转变。
The Trump administration’s scientific agenda has been widely characterized — rightly so — as a war on scientific progress. But, hear me out here: There is more to the story.
This administration’s science policy is being shaped not solely by anti-science ideologues, but also by a motley coalition of players who have distinct criticisms of the status quo and are united by their willingness to part ways with established orthodoxies. They include animal advocates, some of them scientists themselves, who quite reasonably hope to advance science beyond its current dependence on animal experimentation.
Sign up here to explore the big, complicated problems the world faces and the most efficient ways to solve them. Sent twice a week.
Research animals — from mice, to rabbits, to monkeys — still underpin much of medical research. But their usefulness as models for humans has always been limited. As Harvard bioengineer Don Ingber told me last year, “Everyone admits that animal models are suboptimal at best, and highly inaccurate more commonly.” The ethical problems with experimenting on animals are also immense, and meanwhile, a new generation of animal-free research technologies is proliferating, including lab-made organoids, organs-on-chips, and advanced computational modeling.
Following on this line of reasoning, the National Institutes of Health (NIH), chief underwriter of university biomedical research in the US, last year under the leadership of director Jay Bhattacharya announced its intent to prioritize animal-free methods and reduce the use of animals in the science it funds. And, together with a major US biomedical research university, it just took a major step toward that goal.
This week, the board of Oregon Health and Science University (OHSU), which runs the one of nation’s largest university centers for biomedical research on primates, voted unanimously to begin negotiating with the NIH about the agency’s proposal to end experiments on the primates and turn the center into a sanctuary for the animals. Many opponents of animal research hope this can create momentum for a phaseout of experimentation on our primate cousins.
OHSU’s primate research center, one of seven such federally funded centers still running at universities across the country, houses about 5,000 monkeys of various species — about 5 percent of all research monkeys in the US — including rhesus macaques, Japanese macaques, baboons, and squirrel monkeys. As part of the resolution reached this week, the center will stop breeding new monkeys, except as required by current experiments, while it discusses a potential plan with the NIH over the next six months to evolve from a primate breeder and experimentation facility to a sanctuary.
OHSU has been dogged by controversy over conditions for animals there, including dozens of citations for violations of federal animal welfare law over the past few decades. Two monkeys died in 2020 after a worker accidentally placed them in a cage-washing machine, while, in 2023, a newborn monkey was killed after being hit by a falling sliding door, to name a couple examples.
“[OHSU’s] record is one of the worst I’ve seen,” Delcianna Winders, a professor and director of Vermont Law and Graduate School’s Animal Law and Policy Institute, told me. “They just have negligent death after negligent death.” (Disclosure: In 2022, I attended a media fellowship program at Vermont Law and Graduate School.)
At a public meeting on Monday, researchers at the university’s primate center, along with others from the university and members of the general public, fiercely debated the proposal to end research at the center. “Past research in primates might have contributed to the advancement of medicine, but it is evident that the advanced methods now available have rendered it virtually obsolete,” said Michael Metzler, an emergency physician at Pioneer Memorial Hospital in Oregon. “These monkey studies divert funds and attention from the more valuable human-centered studies.”
Supporters of the primate center, meanwhile, condemned the university’s “immediate surrender to a hostile administration over political pressure,” as Cole Baker, a PhD student in biomedical engineering at OHSU, put it at the hearing.
OHSU is no doubt under pressure to cooperate with the NIH, which, as of fiscal year 2023, provided the majority of the university’s research funding, and the White House has shown that it’s perfectly willing to punish universities that don’t comply with its wishes. But calls to close the center predate the Trump administration, and it is hardly just a Republican priority. Oregon’s Democratic governor Tina Kotek has urged the primate center’s closure, citing the example of Harvard University, which closed its own primate research center in 2015 amid controversy over its treatment of monkeys.
Harvard’s decision itself is a noteworthy signal of where medical research is headed. One of the world’s top biomedical research institutions apparently determined — more than a decade ago — that the medical science coming from its primate research center wasn’t worth its continued financial, reputational, and ethical costs.
Debates over the necessity of primate research can be hard to parse. Advocates on either side of the question appear to be speaking different languages, with opponents arguing that animal data tells us very little that’s applicable to humans, and proponents insisting that they couldn’t possibly conduct research into debilitating human diseases without using monkeys.
Thomas Kuhn, the 20th-century historian of science who coined the phrase “paradigm shift,” had a name for such breakdowns in communication: incommensurability. Scientists working within different paradigms can see the same thing and come to radically different conclusions because they are looking at problems through different conceptual lenses.
And scientists are still often siloed, as neuroscientist and Vox contributor Garet Lahvis, a former professor at OHSU who spoke in favor of ending research at the primate center at the hearing this week, pointed out to me. Primates are used in a wide range of research applications, including infectious diseases, neuroscience, psychology, reproductive health, and more, and that very specialization, he pointed out, can make it hard for scientists to take a broader scientific perspective.
Primate research, like most things in science, is the product of path dependency and historical circumstance. In the 1960s, the US created a system of federally funded primate centers, like the one at OHSU. The NIH at the time “thought primate experiments were the future,” Winders told me, and it has shaped the way lots of medical science is practiced to this day.
But today, the sight of caged lab monkeys looks more like a relic of the past.
It now appears beyond doubt that at least some of what primates are used for in US labs is of extremely limited value, particularly research that aims to model complex mental health conditions in humans, like depression, by inducing them in monkeys. Former NIH director Francis Collins acknowledged as much in 2014, when he referenced “the pointlessness of much of the research being conducted on non-human primates” in a private email that was obtained by PETA as part of a lawsuit.
And the primates’ very captivity might make results even less translatable to humans. Lahvis, for example, has argued that extreme confinement in cages stunts the health of lab animals and skews the psychology of monkeys to such a degree that they can hardly be seen as sound proxies for healthy humans.
While proponents of primate research cite its use in human drug development, like therapies for HIV, the mere presence of primate data in the evidence chain for a medical treatment does not prove that that research was indispensable. And given the high moral stakes of research on social, cognitively complex animals, and the substantial opportunity costs of devoting resources and careers to primate labs, merely being sometimes useful does not seem like sufficient justification for subjecting monkeys to lifelong captivity and invasive experiments.
The NIH deserves credit for acting on this perspective. And there is precedent for phasing out research on a class of animals. The federal government a decade ago ended biomedical research on chimpanzees, although other primates are more deeply embedded in such research than chimps were. So, the NIH now faces the challenge of winding down that research enterprise in a way that respects researchers’ careers; building a credible off-ramp to animal-free research tools; and, in its proposal to fund a primate sanctuary, providing some measure of justice for the animals harmed in federally funded science.
That would be no small task for even a normal administration — and for one that has wrecked its credibility with the scientific community, it will be even harder. Consider it a test case for whether the Trump administration can, amid its ruthless cuts to research, contribute to at least one positive paradigm shift in science.
2026-02-13 20:45:00
2025年1月25日，索马里邦特兰防卫部队的士兵们返回前线基地，该基地位于邦特兰的达布达马勒地区，以打击ISIS武装分子。| 《华盛顿邮报》/盖蒂图片社
要点总结
2025年2月3日，特朗普在社交媒体上发布了一篇关于美国在索马里打击ISIS领导人的福克斯新闻文章，同时对国会议员伊尔汗·奥马尔进行了带有种族色彩的侮辱。特朗普对奥马尔的攻击已成为常态，但在国家安全领域，提及空袭行动显得尤为不寻常。这是他一年来首次在社交媒体上提到对索马里的军事行动，尽管该国在这一时期遭受了比以往任何时期更多的空袭。在特朗普的两个任期内，美国对索马里的空袭都相对低调，几乎没有公开解释。虽然特朗普并非唯一在索马里进行空袭的总统，但他的行动规模远超以往。根据新美国（New America）的数据，2025年美国在索马里进行了125次空袭和一次地面突袭，而拜登总统整个任期仅进行了51次行动。2026年，美国在索马里已执行了28次行动，超过了任何非特朗普总统任期内的完整一年。据估计，在特朗普第二个任期的空袭中，死亡人数在172至359人之间，但美国非洲司令部（AFRICOM）自去年4月起未公布伤亡数据，因此实际数字可能更高，且难以确定有多少是平民。相比之下，美国在加勒比海和东太平洋地区针对所谓贩毒船只的行动仅进行了34次空袭。此外，美国在索马里的空袭数量甚至超过了2010年奥巴马政府在巴基斯坦的无人机战争高峰期。
特朗普和高级官员很少提及他们在索马里的空袭行动，有时甚至表现出对这种长期行动的回避。事实上，他唯一一次在军事背景下提到索马里，是为了炫耀自己没有参与该国事务。他说：“只有在最近几十年，政治家们才相信我们的职责是去警察肯尼亚和索马里的偏远地区，而美国却正遭受内部入侵。”
为何美国对索马里进行如此频繁的空袭？
有多个因素在起作用：对索马里在全球恐怖主义复兴中作用的担忧、放宽了对平民的保护规则，以及一个几乎可以自动运行的反恐战争机器。
索马里自1990年代初以来一直处于内战和人道主义危机之中，美国也参与了近三十年的反恐行动。1993年“黑鹰坠落”事件中，18名美国海军陆战队员在摩加迪沙阵亡，这是美国自越南战争以来最严重的军事损失。9/11事件后，美国开始对索马里的恐怖分子进行空袭和特种部队突袭，特别是针对与基地组织有关联的索马里青年党（Al-Shabab）。该组织曾控制索马里大部分领土，包括首都摩加迪沙。如今，它虽然分散，但仍活跃于索马里各地，并对索马里政府和外国部队发动致命袭击。
自特朗普重返白宫以来，美国对索马里的空袭不仅针对青年党，还越来越多地打击ISIS索马里分支。该分支由青年党叛徒创建，与ISIS是死敌。近年来，ISIS在全球范围内发动了多起高调袭击，专家认为索马里分支在策划这些行动中发挥了关键作用。美国非洲司令部的第二号人物约翰·布伦南（John Brennan）最近告诉福克斯新闻，美国在索马里的反ISIS行动旨在“破坏针对美国本土和欧洲的阴谋”。他声称，索马里ISIS领导人阿卜杜勒·卡迪尔·穆民（Abdalqadir Mumin）实际上是“全球ISIS网络的哈里发”，并从索马里北部的戈利斯山脉据点指挥全球活动。然而，这一说法受到质疑，许多反恐专家并不认为穆民是全球哈里发。该组织从未正式宣布他获得这一头衔，而且他并非阿拉伯人，也不声称是先知穆罕默德的后裔，因此不太可能成为哈里发。
尽管穆民是否为哈里发存在争议，但专家普遍认为，索马里ISIS分支虽然在该国本身只有少数地面战士（据联合国估计，最多200-300人），但已成为全球ISIS网络中最重要的分支之一，负责筹款、融资和招募。反恐分析师科林·克拉克（Colin Clarke）表示：“无论穆民是否是头目，他在全球ISIS网络中都极具影响力，因此是一个高价值目标。从战场上移除他是一个值得的目标，他是全球网络中的关键一环。”
大多数针对ISIS的空袭发生在索马里北部，与邦特兰政府合作进行。与此同时，针对青年党的行动仍在南部持续进行，最近一次打击行动就发生在上周。青年党在索马里国内仍控制着大量领土，并在2025年初短暂占领了距离首都摩加迪沙仅30公里的政府建筑。然而，与ISIS相比，青年党在全球范围内的影响力较小。
特朗普改变了空袭规则
一个可能的解释是，特朗普对索马里的空袭行动并不特别关注。根据政府授权，白宫可能不需要对每次空袭进行审批。这更像是特朗普允许后9/11时代的反恐战争继续进行，而且比之前更加缺乏监督。由于缺乏公众关注，政府也不必详细解释其行动目标或成本。
新美国（New America）的数据表明，特朗普任期内的空袭数量也大幅增加。2017年，特朗普放宽了防止平民伤亡的规则，使AFRICOM能够更自由地选择目标。在重返白宫后，他再次放宽了这些限制，这可能是导致索马里空袭增加的主要原因。
反恐行动“自动运行”
新的空袭规则解释了行动规模的扩大，但并未说明为何选择在索马里进行。一个可能的解释是，索马里是“最方便的目标”。美国的反恐行动似乎存在路径依赖，即在过去的反恐行动中形成的惯性。2024年，美国从尼日尔撤军，使得在萨赫勒地区（非洲西部）打击恐怖分子的资源减少。此外，美国两党都不愿重返阿富汗，而反恐行动在叙利亚（美国仍有驻军）和索马里（有长期合作和邻近的美军基地）仍在继续。
特朗普与索马里的关系
特朗普虽然很少公开谈论索马里的空袭行动，但经常提及索马里本身，尤其是在美国移民官员在明尼苏达州（索马里移民社区聚集地）展开争议性且暴力的打击行动后。该行动表面上是为了打击索马里企业涉嫌的社会福利欺诈。在达沃斯世界经济论坛上，特朗普称索马里人是“低智商的人”，并称索马里“不是一个国家”。副总统JD·范斯（JD Vance）也提到美国存在“索马里问题”。
美国与索马里政府的关系最近也紧张。美国削减对外援助严重破坏了索马里的医疗系统，使许多儿童无法获得粮食援助。美国上个月曾短暂暂停对索马里的所有粮食援助，因指控当地官员劫持了世界粮食计划署的仓库。
尽管特朗普对索马里的敌意可能与大规模空袭有关，但这种联系似乎不太可能。空袭行动是在与索马里政府和邦特兰当局密切合作下进行的，而周围的政局并未影响行动的进行。索马里驻美大使达希尔·哈桑·阿比（Dahir Hassan Abdi）表示：“索马里政府不会将政治言论视为政策的替代品。美国仍是安全合作的关键伙伴，索马里则专注于推进共同目标的实际协调。”
空袭是否有效？
大使达希尔认为，美国的支持帮助索马里的武装力量削弱了恐怖分子，恢复了一定的稳定。他指出：“恐怖分子攻击主要城市和政府部队的能力下降，为摩加迪沙的和平居民创造了自由参与地方选举的条件，这是五十年来首次。”
然而，美国空袭在以往的反叛乱行动中效果不佳，反恐专家乔舒亚·梅塞维（Joshua Meservey）表示，空袭未必能成为解决索马里问题的关键。他指出：“索马里的核心问题是缺乏有能力和合法的本地治理。”
此外，有官员担心空袭可能导致更多平民伤亡，从而加剧索马里民众对政府和美国支持者的不满，甚至催生更多恐怖分子。这些动态并非特朗普独有，而是美国反恐行动长期存在的问题。在这一背景下，特朗普不愿亲自公开谈论行动也情有可原。他更倾向于快速、决定性的胜利，而索马里的行动则是一个持续多年、成效难以界定的战争。在前任建立的体系中，只需他的默许，就能维持这场遥远的战争。
On February 3, President Donald Trump posted a Fox News article about a US strike targeting ISIS leaders in Somalia, along with an inflammatory insult aimed at Congresswoman Ilhan Omar, who arrived in the US as a refugee from the country.
Trump taking a racist dig at Omar has become routine. But in national security circles, the mention of the strikes stood out as unusual.
The post, along with a similar one the day before, was the first time in a year that the president’s account had mentioned his military campaign in Somalia, despite bombing the country more than any other in the same period.
In both his terms as president, Trump has quietly overseen a massive escalation of airstrikes in Somalia with little public explanation. And while the president is not unique in ordering strikes there — the military has been enmeshed in the Horn of Africa country’s conflicts since the early 1990s — his campaign is simply on another level, as shown by data compiled by New America.
In 2025, the US carried out 125 airstrikes and one ground raid in Somalia, compared to 51 operations during Joe Biden’s whole presidency. Already, in 2026, the US has carried out 28 operations, more than any full year under a non-Trump president. Between 172 and 359 people have been killed in Trump’s second term strikes, though David Sterman, a counterterrorism analyst at New America, notes that US Africa Command (AFRICOM) has not been reporting casualty estimates from these strikes since April of last year, meaning that the real numbers are likely much higher, and it’s difficult to know how many were civilians.
By comparison, the high-profile US campaign against alleged drug trafficking boats in the Caribbean and Eastern Pacific since last fall has consisted of just 34 strikes.
Or, another point of comparison: The US carried out more strikes in Somalia last year than it did in Pakistan in 2010 — the height of the Obama administration’s drone war — when these tactics were a matter of major controversy and national debate.
Not only do Trump and senior officials rarely talk about the fact that they are waging an air war in Somalia that rivals the height of the Global War on Terror, the statements the president does make sometimes suggest an aversion to exactly this type of open-ended campaign. In fact, one of the rare times he’s brought up Somalia in a military context was to boast about not getting involved in the country.
“Only in recent decades did politicians somehow come to believe that our job is to police the far reaches of Kenya and Somalia, while America is under invasion from within,” Trump told a group of US military leaders at a gathering in Quantico last fall.
So, why is the US bombing Somalia so much, and why isn’t anyone talking about it?
There appear to be several factors in play: a genuine growing concern about Somalia’s role in a global resurgence of jihadist terrorism, a loosening of rules protecting civilians that allows for more strikes, and a post-9/11 war machine that can operate almost automatically without the president’s personal attention.
Somalia has been in a state of civil conflict and humanitarian crisis since the early 1990s, and the US has been involved for almost that long. The deaths of 18 US Marines in the infamous “Black Hawk Down” incident in Mogadishu in 1993 was the worst loss of life for the US military since Vietnam at that time.
After 9/11, the US began to target militants in the country with airstrikes and special forces raids, particularly the newly emerged al-Qaida-linked movement known as Al-Shabab. At various points, al-Shabab has controlled large swaths of Somali territory, including parts of the capital, Mogadishu. Today, it is a more dispersed movement but is still active in much of the country and continues to carry out deadly attacks against the Somali government and foreign troops in the country.
The war against al-Shabab has continued with various levels of intensity across several administrations, but the biggest change since Trump returned to office, experts say, is that, in addition to al-Shabaab, the strikes reported by AFRICOM are increasingly targeting ISIS’s affiliate in the country. (ISIS’s Somali “province” was founded by al-Shabab defectors, and the groups are sworn enemies.) The Islamic State has carried out a number of high-profile global attacks recently, and experts believe the Somali affiliate is playing a key role in facilitating those plots.
Lt. Gen. John Brennan, the second-highest-ranking officer at AFRICOM, recently told Fox News that the stepped up anti-ISIS campaign in Somalia is in order to disrupt plots “against the United States homeland as well as Europe.”
Notably, Brennan also claimed that Abdalqadir Mumin, the leader of ISIS in Somalia, is in fact “the caliph — absolute leader — of the global ISIS network” and is directing ISIS’s global activities from his hideout in the Golis Mountains.
That assertion, which first emerged after Mumin was unsuccessfully targeted in a 2024 strike, is contested. Many terrorism experts do not believe Mumin is the global caliph. The group never announced he’d been given the title, and, as a non-Arab who doesn’t claim descent from the Prophet Mohammed, he would make an unusual choice.
Caliph or no, there is growing consensus among terrorism experts that the Somali affiliate, which has relatively few fighters on the ground in Somalia itself (as few as 200-300 according to UN estimates), has become one of global ISIS’s most significant global affiliates, playing a key role in fundraising, financing, and recruiting.
“Whether Mumin is the head or not, he’s extremely influential within the Islamic State’s global network, so he’s a high-value target, clearly. Removing him from the battlefield is a worthwhile objective; he’s a vital cog in the global enterprise,” said Colin Clarke, terrorism analyst and executive director of the Soufan Center.
The vast majority of the strikes against ISIS have been in northern Somalia in cooperation with the security forces of the semi-autonomous Puntland state. But the campaign against al-Shabab in southern Somalia continues apace, as well; a strike against the group took place just last week. Al-Shabab is more formidable with Somalia and has carried out high-profile attacks outside of it — mainly in East African countries that have sent troops to Somalia — but it has less of a global reach than ISIS. Some worry that could be changing. In 2024, US intelligence agencies learned of discussions about a weapons deal between al-Shabab and the Houthis, the Iran-backed militant group across the Red Sea in Yemen that Trump has previously targeted, though it’s unclear if anything came of those talks.
So, it’s not surprising that the US would pay attention to Somalia as part of an overall global campaign against jihadist terrorism. US officials have even evoked the Israeli phrase “mowing the grass” to describe their goal in Somalia: keep militant groups degraded to prevent them from becoming too much of a threat.
But this doesn’t fully explain the shift. Somalia is hardly the only country where these groups are a threat. The Afghan ISIS affiliate, ISIS-Khorasan, has carried out major recent attacks in that country and abroad. The US has not carried out a publicly reported military operation in the country since the drone strike that killed al-Qaeda leader Ayman al-Zawahiri in 2022.
Meanwhile, the epicenter of global terrorist violence, accounting for more than half of all deaths, is West Africa’s Sahel Region — particularly countries like Mali, Niger, and Burkina Faso, rather than East Africa. Mali is under literal siege from the local al-Qaeda affiliate, forcing the government to ration fuel. But with the exception of Trump’s Christmas Day bombing of Nigeria, which appears to have been a one-off, the US has not seemed particularly interested in West African jihadist groups.
So, why Somalia in particular?
The simplest theory for why Trump doesn’t speak much about his administration’s most extensive military operation is that he’s not particularly involved with it. Under the authorities the administration has granted, the White House most likely doesn’t need to sign off on individual strikes.
It’s less of an example of a Trump policy than him allowing one of the remaining vestiges of the post-9/11 war on terror to continue — with even less oversight than before. The lack of public attention on the operation has meant the administration is under little pressure to fully explain its goals or justify its costs.
As New America’s numbers show, the number of strikes in Somalia also grew dramatically in Trump’s first term. One big reason: In 2017, Trump relaxed rules meant to prevent civilian casualties, giving AFRICOM wider latitude to go after targets as it saw fit. And after returning to office a second time, Trump again relaxed these limits, which appears to have been the main factor leading to the uptick in operations in Somalia.
“It’s very clear that they’re operating under significantly expanded authorities for strikes again,” said Sterman, who tracks reports of strikes in Somalia for New America.
The most extensive public discussion of the shift from a Trump administration official came last July from Sebastian Gorka, the National Security Council’s Director for Counterterrorism, during an appearance at the Foundation for the Defense of Democracies in Washington.
“You may not be aware of it in the broader universe, but we are stacking [jihadis] like cord wood,” Gorka declared.
Gorka said that, shortly after taking office, he was told by intelligence and defense officials that “we’re not allowed to kill bad guys” under Biden’s targeting review rules. Gorka claims that on “day eight of the administration” he presented the president with evidence of an “ISIS jihadi running freely around a terror compound, a cave system in Northern Somalia,” who had been tracked for years. Trump quickly signed off on the order to kill the jihadi, after which Gorka watched on a screen as the man was turned into “red mist.”
This strike, in February, 2025, was one of the only times Trump has tweeted about the campaign, calling it a “message to ISIS and all others who would attack Americans is that ‘WE WILL FIND YOU, AND WE WILL KILL YOU!’”
Under the more relaxed targeting standards that have been in place since the summer, this consultation with Trump would not have even been necessary. The rationale for these strikes has also subtly changed. According to Sterman, strikes under previous administrations, including the first Trump term, had often been justified as “collective self-defense” operations, meaning the US was responding to an attack on either US personnel or its Somali allies. That language appears less often now. It’s also possible Biden may in fact have set the stage for this new offensive by preemptively signing off on the targeting of about a dozen Shabab leaders, meaning AFRICOM may feel more comfortable calling in the drones in the absence of a pressing threat.
The new targeting standards help account for the scale of the bombing but not why it’s happening in Somalia, specifically, as opposed to other countries with Islamic militants that the US has targeted in the past.
One possible explanation is that it’s the most convenient target. There appears to be an element of path dependence in the way counterterrorism is carried out by the US today. Having withdrawn troops from what was once a major counterterrorism hub in Niger in 2024, the US has fewer resources for combating jihadists in the Sahel than it once did. There’s little appetite in either US party for a return to Afghanistan after Biden’s ugly withdrawal in 2021.
By contrast, anti-ISIS operations have continued in Syria, where the US still has a troop presence — though that could be ending soon — and in Somalia, where there’s a history of these operations and cooperation with local forces, as well as US troops stationed nearby in Djibouti, Kenya, and Somalia itself.
“It seems like it’s on autopilot,” said Tibor Nagy, a veteran US diplomat who served as under secretary of State for African Affairs during Trump’s first term. “It’s easier to keep doing something because there’s the institutional bureaucracy in place to keep supporting it.”
Asked about the reason for the uptick in strikes, a Department of Defense spokesperson speaking on background told Vox: “Our strategic approach to countering terrorism in Africa relies on trusted partnerships and collaboration grounded in and through shared security interests. The cadence in conducting airstrikes in Somalia reflects that strategy, enabled by the administration’s policy to empower commanders to protect the U.S. homeland and citizens abroad.”
Trump may not talk about the air campaign in Somalia, but he has been talking about Somalia itself quite a bit, particularly since US immigration officials began a contentious and violent crackdown in Minnesota — home of a large Somali immigrant community — ostensibly motivated by cases of social services fraud by individual Somali owned businesses.
In a recent high-profile speech at the World Economic Forum in Davos, Trump said he had always thought of Somalis as “low-IQ people” until seeing the scale of the fraud and called Somalia “not a country.” Vice President JD Vance has described the United States as having a “Somali problem.”
Relations between the US and Somali government have also been strained lately. US foreign aid cuts have devastated the country’s health care system and left children in much of the country without food aid. The US briefly suspended all food aid to Somalia last month over allegations that local officials had seized a World Food Program warehouse.
It’s tempting to wonder if Trump’s general enmity toward Somalia is related at all to his massive bombing campaign in the country, but that seems unlikely.
The air campaign is conducted in close coordination with the government of Somalia and Puntland authorities. If anything, what’s notable is that the surrounding politics haven’t disrupted the campaign.
“Somalia’s government does not treat political statements as a substitute for policy,” Somalia ambassador to the United States, Dahir Hassan Abdi, responded by email when asked about Trump’s comments. “The United States remains a critical partner in security cooperation, and Somalia remains focused on practical coordination that advances shared goals.”
Dahir, the Somalia ambassador, argued that US support has allowed its forces to put the jihadists on the back heel and restore a bit of stability.
“The degradation of terrorists’ ability to attack major cities and government forces have created the conditions for peace-loving residents of Mogadishu to freely participate in local elections on December 25, 2025, for the first time in five decades,” he said in an emailed statement from the embassy.
Those local elections, held last year, were billed as a kind of rehearsal for national elections — which are planned for this year — despite concerns about violence and instability.
But while it’s generally agreed that Shabab no longer poses the existential threat to the Somali state that it did in the past, it still controls a significant amount of territory outside the capital and, early last year, briefly captured government buildings just 30 kilometers from Mogadishu.
“The Somali government’s in a decent enough position that it’s not about to fall,” said Omar Mahmood, a senior analyst at the International Crisis Group, speaking by phone from Somalia. “But the question is always, how stable is the Somali government.” Mahmood noted with an upcoming contested election, an international peacekeeping mission underfunded, and the US withdrawing much of its non-military support, the concern is that some of these underlying gains could unravel. That would allow al-Shabab to advance.”
Air strikes don’t have a great record as a counterinsurgency tool in past conflicts, and Joshua Meservey, a senior fellow at the Hudson Institute and author of a book on al-Shabaab, was skeptical that they would make the key difference in this place. “The core problem in Somalia is that there is a lack of competent, legitimate local governance in the country,” he said. “If you do not have that, you will never successfully eradicate these groups.”
Former officials who spoke with Vox also expressed concerns that civilian casualties — about which we have little publicly available information — could turn more Somalis against the government and its US backers and potentially create more militants. None of these dynamics are particular to Trump. If anything, the Somalia campaign is an illustration that the militant groups were the primary focus of US national security for the 20 years after 9/11 have not gone away, even if we don’t talk about them as much, and that efforts to combat them are still slow-going and legally murky.
In that context, it makes sense Trump is reluctant to bring the operations up himself. The president surely approves of the killing of senior al-Shabab and ISIS leaders, but this is a leader who likes quick, decisive, and overwhelming victories. A now-decades-old operation whose success is difficult to define is not that. In the machine built by his predecessors, though, all that’s needed to maintain a simmering war thousands of miles away is his tacit consent.
2026-02-13 20:15:00
在允许AI机器人使用其内容进行训练后，关于维基百科以人类为主导身份的未来引发了诸多疑问。如果你曾在课堂上使用过电脑，很可能听过这样的告诫：不要相信维基百科。其理由是任何人都可以修改维基百科的页面。虽然这在很大程度上属实，但某些页面会因遭受大量滥用或破坏而被锁定。然而，学校对维基百科不可靠或充斥错误信息的观念被夸大了。这个众包在线百科全书依赖于一群被称为“维基人”的志愿者，他们遵循严格的编辑流程。每篇文章底部都有参考文献，每个条目都有公开的“讨论页面”，允许编辑者讨论修改并达成共识。此外，该网站还拥有高效的监控系统，由可靠的编辑和经过维基百科认证的机器人实时监督内容。
“我们被教导不要在课堂上使用维基百科，这让我感到很沮丧，因为我们并没有真正学会如何使用它，”22岁的内容创作者Dean告诉Vox。去年12月，Dean在TikTok上发布了一条呼吁粉丝使用维基百科的视频，强调了它在信息泛滥时代的重要性。他和其他许多社交媒体上的创作者和用户一样，正在重新发现这个25年历史平台的可信度和价值，尤其是在一些有时存在错误的AI聊天机器人崛起的同时。
例如，BBC在2024年12月进行的一项研究发现，像OpenAI的ChatGPT和微软的CoPilot这样的大型AI模型在被提示时会错误地总结新闻。而《卫报》在2026年1月的一项调查则发现，谷歌的AI摘要功能正在向用户展示错误和误导性的医疗信息，这可能危及他们的健康。所有这些都使得一个由人类生成并严格监控的知识平台显得更加有吸引力。
如果公众意识到维基百科的重要性，维基媒体基金会（该基金会资助和支持维基百科）在2025年筹集了高达1.84亿美元的资金，比前一年增加了400万美元。与此同时，《华盛顿邮报》在2025年8月报道说，“可疑的编辑，甚至是完全新的文章，带有错误、虚构的引用和其他AI生成写作的特征，不断出现在这个免费的在线百科全书上”，迫使人类编辑去发现并纠正这些错误。如今，维基百科正在直接与许多用户认为其抗衡的大型语言模型合作。2025年1月，该组织宣布了一批新的科技公司，这些公司将使用维基百科的付费产品Wikipedia Enterprise来训练其AI模型。这并不是一个前所未有的举措，但它引发了对这个早期互联网标志性平台未来发展的担忧。在AI对互联网的控制日益增强的情况下，维基百科如何保持其以人类为主导的身份？
尽管维基百科在线上仍受到一定的怀旧情怀青睐，但它也面临着与其他数字出版商相同的可见度下降问题。去年10月，该组织报告称，其每月的人类页面浏览量相比2024年下降了约8%，并将其归因于人们越来越多地使用生成式AI来获取信息，而这些AI工具通常会直接从维基百科获取信息。此外，人们在需要信息时也更倾向于在社交媒体上搜索。据研究显示，目前有五分之一的美国人经常从TikTok获取新闻，而使用ChatGPT的美国人数量自2023年以来已经翻了一番。
但这并不意味着维基百科已经过时。它仍然是谷歌搜索结果和AI摘要中列出的首要信息来源。过去十年间，维基百科的文章被浏览了1900亿次，2025年成为全球第九大访问量最高的网站。
在TikTok上，维基百科的怀旧情怀似乎仍然存在。像@tldrwikipedia和@depthsofwikipedia这样的流行Instagram账号在过去几年里成功地利用了这种怀旧情绪。后者由Annie Rauwerda运营，展示了维基百科一些更具体和奇特的页面，拥有160万粉丝。在TikTok上，也有专门的维基百科粉丝群体，他们传播该网站及其应用程序的福音，并分享自己浏览随机文章的喜爱之情。去年秋天，一位不愿透露姓氏的22岁毕业生Chisom发布了一条TikTok视频，说她“毫无保留地买了一顶维基百科帽子”，该视频获得了百万次观看和大量正面评论。Chisom表示，她以前认为维基百科不可靠，直到一位高中老师展示了该网站的监控系统如何有效运作，以及如何快速进行实时更正。现在，她说自己成了“深入维基百科的爱好者”，并认为维基百科比谷歌的AI摘要更易于使用。“我确实使用得更多了，”她说，“以前我用谷歌，他们会提供一些名人的简要信息，比如他们结婚的对象、孩子等。但自从他们开始使用AI摘要，这对我帮助不大。”
尽管AI的威胁依然存在，但人类仍然为维基百科的未来提供了希望。尽管维基百科在线上重新获得了关注，但随着AI逐渐渗透到我们日常生活的各个方面，尤其是维基百科与AI公司建立合作关系，这似乎与其以人类为中心的原则相冲突。大型语言模型已经使用维基百科一段时间了，而且通常是未经允许的，这对维基百科造成了很大的负担。科技记者Stephen Harrison在Slate上多年报道维基百科，他对维基百科与AI公司的合作表示，“这是科技公司对维基百科等项目的一种认可，认为它们的长期发展依赖于这些项目。”他更担心的是像埃隆·马斯克这样的公众人物对维基百科发起的政治攻击。去年，马斯克批评并呼吁削减维基百科的资金，因为他的条目被更新以指出他在特朗普就职典礼上做出的一个手势，被广泛解读为纳粹礼。随后，他推出了竞争对手网站Grokipedia，其条目由他的公司xAI编辑。
Harrison还担心，如果用户主要通过AI摘要来接触维基百科，他们可能会逐渐忘记这个网站。维基百科志愿者Hannah Clover自2018年以来一直参与维基百科的工作，她对AI的影响也有自己的担忧。她并不认为AI会取代人类编辑，但担心AI的普及会使信息来源变得更加难以追踪。“我担心的是，我们引用的许多来源在未来可能会变得不可靠，”Clover说，“我们有一个长期的来源列表，有时一些原本可靠的来源会因为突然开始发布AI生成的垃圾信息而变得不可靠。”
这些与AI的合作表明，维基百科仍然是一个极其重要的知识库。但最终，维基百科的存续仍取决于那些热爱它的人。Clover承认，许多年轻人因为经济压力而没有时间或精力成为维基百科的编辑，但这并不是“缺乏兴趣”的表现。Harrison则认为，像“深度维基百科”这样的独立创作者对于维基百科品牌的延续至关重要。“社交媒体影响者依赖维基百科作为他们知识的隐形基础，”他说。目前，TikTok上所有关于“旧互联网”的怀旧内容给了他一些希望，认为维基百科可能会迎来复兴。“我成长于维基百科被视为互联网‘荒野’的时代，”他说，“维基百科在很多方面已经成为一个充满故事的机构，人们对其怀有怀旧和喜爱之情。”
If you grew up with computers in your classroom, there’s a good chance you heard this instruction before starting a research paper: Don’t trust Wikipedia.
The reasoning? Anyone can go in and make changes to a Wikipedia page. This is mostly true, though pages that are subject to a high amount of abuse or vandalism can be locked. However, the notion that the website is unreliable or a playground of misinformation has been overstated in schools. The crowdsourced online encyclopedia relies on a community of volunteers, known as “Wikipedians,” who adhere to a rigorous editing process. Citations are available at the bottom of each article, and public-facing “talk pages” attached to every entry allow editors to discuss changes and try to reach consensus. And the site has an efficient monitoring system, with reputable editors and Wikipedia-approved bots watching entries in real time.
“The fact that we were all told not to use it in school is really frustrating because we just weren’t taught how to actually use it,” Dean, a 22-year-old content creator, told Vox.
Last December, Dean posted a TikTok urging his followers to utilize Wikipedia, emphasizing its importance in an era of rampant misinformation. He’s like many other creators and users on social media who are discovering the credibility and value of the 25-year-old platform in the same moment that sometimes-faulty AI chatbots are ascendant. For example, research conducted by the BBC in December 2024 found that major AI models like OpenAI’s ChatGPT and Microsoft’s CoPilot inaccurately summarize news when prompted, and a Guardian investigation in January 2026 found that Google’s AI Overview was showing users false and misleading medical information that put their health at risk.
All of this makes a knowledge platform that’s human-generated and rigorously monitored look pretty appealing. If there’s any indication that the public understands its necessity, the Wikimedia Foundation, which funds and supports Wikipedia, raised a staggering $184 million in 2025, a $4 million increase from the previous year.
At the same time, the Washington Post reported in August 2025 that “suspicious edits, and even entirely new articles, with errors, made-up citations and other hallmarks of AI-generated writing keep popping up on the free online encyclopedia,” forcing human editors to find and fix them. And Wikipedia is now working directly with the large language models that many users see it as counterbalancing. In January, the organization announced a new batch of tech companies that will train their AI models using Wikipedia Enterprise, a paid product allowing partners to access its content at scale. This isn’t an unprecedented move, but it raises concerns about the future of the early-internet staple. How will it maintain its human-powered identity amid AI’s chokehold on the internet?
Wikipedia might have had a big year for fundraising, but it’s faced the same struggles with visibility as other digital publishers. Last October, the organization reported that its monthly human page views had seen a roughly 8 percent decline compared to 2024 and attributed it to the uptick in people using generative AI — which, again, uses Wikipedia as a source and provides the info directly to users — and searching on social media when they need information. (Research has shown that one in five Americans regularly gets their news from TikTok now, while the number of Americans using ChatGPT has already doubled since 2023.)
That’s not to say Wikipedia’s gone out of fashion. It remains a top source listed in Google search results and AI summaries. Over the past decade, its articles have been viewed a total of 1.9 trillion times. It was the ninth most-visited website in the world in 2025.
There also seems to be a niche, nostalgic appeal to Wikipedia that persists online. It’s something that popular Instagram accounts like @tldrwikipedia and @depthsofwikipedia have been able to capitalize on over the past few years. The latter, run by Annie Rauwerda, features screenshots of the site’s more specific and bizarre pages and boasts 1.6 million followers.
On TikTok, there’s a dedicated Wikipedia fandom, with users spreading the gospel of the website (and its app) and sharing their affinity for browsing random articles. Last fall, Chisom, a recent grad and substitute teacher who prefers not to share her last name online for privacy reasons, posted a TikTok saying she “unironically bought a Wikipedia hat.” It received a million views and loads of positive comments.
Chisom, 22, told Vox she grew up believing that Wikipedia was unreliable until a 10th-grade teacher demonstrated how well the site’s monitoring system works and how quickly corrections are made in real-time. Now, she said, she’s become “rabbit-hole Wikipedia girl” and finds it much more user-friendly than Google’s AI overview.
“I definitely use it more,” she said. “I used to use Google, and they would have a little summary of a celebrity — who they’re married to, their kids. But since they started doing the whole AI summary thing, that’s so unhelpful to me.”
Despite renewed enthusiasm for Wikipedia online, the future of the site seems tenuous as AI creeps into more aspects of our everyday lives, and, especially, because seeing the website establish relationships with AI companies feels at odds with its human-first principles. Large language models have been using Wikipedia for a while now, famously without their permission and at a high cost to the site.
Tech journalist Stephen Harrison, who covered Wikipedia on Slate for years, told Vox that he sees the LLM partnerships as “recognition” by tech companies that “their long-term future depends on nurturing projects like Wikipedia.” He’s more concerned about the political attacks the platform has faced recently from people like Elon Musk. (Last year, Musk criticized and called to defund Wikipedia after his entry was updated to note a gesture he made during Trump’s inauguration that was widely interpreted as a Nazi salute. He’s since launched the rival website Grokipedia, with entries edited by his company xAI.) Harrison is also concerned about internet users “forgetting” about Wikipedia if they’re mainly consuming the site’s content through AI summaries.
Hannah Clover, a Wikipedian who has been working with the site since 2018, told Vox her concerns about AI’s impact are a bit less obvious. It’s not that she believes AI will ever replace human editors, but that its prominence will make sourcing harder.
“I worry about it more in the sense that a lot of the sources that we cite might become unreliable in the future,” Clover, 23, said. “We have a perennial sources list, and sometimes you have sources that were previously reliable that become unreliable because they start publishing AI slop out of nowhere.”
These AI deals show that Wikipedia is still an extremely critical knowledge base. But it will inevitably be up to the humans who love it to keep the site going. Clover acknowledges that a lot of young people struggling to pay their bills may not have the time or energy to become Wikipedians who edit the site, but that’s “not for a lack of interest.” Harrison, meanwhile, sees independent creators, like Depths of Wikipedia, as crucial in keeping Wikipedia’s brand alive. “Social media influencers rely on Wikipedia as a sort of invisible foundation for their knowledge,” he said. For now, all the “old internet” nostalgia on TikTok gives him some hope for a revival.
“I grew up when Wikipedia was considered the Wild West of the internet,” he said. “It’s really remarkable how Wikipedia has, in a lot of ways, become this storied institution that people have all these feelings of nostalgia and affection toward.”
2026-02-13 19:45:00
在2026年2月6日，特朗普乘坐空军一号前往佛罗里达州棕榈滩途中与记者合影。| Samuel Corum/Getty Images 在他的第二个总统任期刚刚过去一年之际，特朗普尚未将美国变成一个纯粹的极权主义噩梦，但这并非因为他没有努力。2025年1月，如果问自由派人士，他们可能会预测特朗普政府在最坏情况下会采取以下行动：• 对政治对手提出虚假指控，包括民主党立法者；• 强迫电视网络取消批评喜剧演员的节目；• 部署军队到民主党控制的城市以镇压对其政策的抗议；• 赋予移民执法官员无限制的权力，让他们可以杀害美国公民。特朗普已经实施了这些措施，但每一次都遭遇了司法界和民间社会的抵制，这些抵制要么挫败了他的独裁企图，要么至少对其进行了限制。本周，这种动态再次显现，特朗普提出了一些夸张的独裁计划，但美国公民和宪法秩序迫使他退缩。
尽管特朗普政府试图起诉六名民主党议员，指控他们犯有“煽动行为”，但最终未能成功。这些议员在去年秋天发布了一段视频，警告美国军方和情报机构人员，“对宪法的威胁不仅来自国外，也来自国内”。这些议员本身都是前军人或情报人员，提醒同行必须拒绝非法命令。他们并未明确质疑任何具体军事行动的合法性，但该视频发布几个月后，国防部长皮特·海格塞思被指控曾下令对一艘委内瑞拉渔船上的人员进行“杀光”——这显然是非法的。有人可能难以理解这些民主党人的言论为何违法：如何能构成非法，一个当选官员鼓励遵守宪法或《日内瓦公约》？然而，特朗普却认为这不仅违法，而且是死罪。他在Truth Social上宣称民主党人犯有“叛乱行为”，并可判处死刑，还转发了一条支持者呼吁“把他们吊死，乔治·华盛顿也会这么做！”的帖子。在这一届政府中，司法部似乎首先效忠于总统的意愿，而非美国法律。因此，本周华盛顿的美国检察官办公室试图起诉所谓的“叛乱六人”并不令人意外，但令人惊讶的是，司法部最终未能成功。大陪审团通常会支持起诉方，因为他们只听取政府一方的说法，只需判断是否有“合理依据”认为被指控者可能犯有罪行。然而，该政府未能达到这一低标准。这反映了更广泛的趋势：司法部试图起诉特朗普的其他对手，如前FBI局长詹姆斯·科米和纽约州检察长丽蒂西亚·詹姆斯，也均被法官和大陪审团驳回。尽管对美联储主席杰罗姆·鲍威尔的刑事调查结果尚不确定，但其意图迫使鲍威尔提前辞职的计划显然失败了。
本周还有一项更安静的迹象表明民主制度的韧性。2025年，特朗普政府多次派遣联邦国民警卫队部队前往民主党控制的城市，以镇压或威慑其不喜欢的抗议活动。在6月的一份备忘录中，特朗普授权警卫队部署到任何“发生或可能发生的针对移民执法局（ICE）职能的抗议活动”地点。换句话说，他声称有权在任何美国城市派遣军队，只要那里有反对他移民政策的示威活动。这违反了美国民主制度的基本规范，即维护民事法律的责任应由文职官员承担，除非在最极端的情况下。幸运的是，该政府对国民警卫队的滥用引发了司法界的批评，包括以保守派为主的最高法院。据《华盛顿邮报》周三报道，特朗普政府已从美国各城市撤回所有联邦国民警卫队部队。
美国民主制度并非没有斗争，但其抵抗特朗普对民主的攻击比人们想象的要顽强得多。在过去一年中，特朗普政府多次试图实施令人不安的独裁行为，但都未能成功。例如，联邦通信委员会试图迫使迪士尼停播吉米·金姆尔的节目，因为其笑话不受欢迎；但娱乐工会和迪士尼+的订阅用户威胁要抵制该公司，最终迫使迪士尼恢复了金姆尔的节目。当明尼苏达州边境巡逻队向抗议者亚历克斯·普雷蒂的倒地身体开枪时，特朗普政府试图污蔑受害者并赞扬凶手，但公民和两党官员要求调查，最终司法部不得不进行调查。此外，对普雷蒂遇害的舆论反弹迫使政府减少了在明尼苏达州前所未有的驱逐行动，该行动曾导致居民的常规民事权利被侵犯。
白宫对移民的正当程序权利发动了公开战争，取得了相当大的成功。然而，其最严重的侵犯之一——非法将基尔马尔·阿布雷戈·加西亚驱逐到一个以折磨囚犯而臭名昭著的外国监狱——却遭到了联邦法官的抵制，最终政府不得不将加西亚送回美国。同样，特朗普政府试图因一篇他不同意的社论而驱逐塔夫茨大学的一名学生，但最终在法庭上败诉。最重要的是，尽管特朗普表现出操纵选举的意图，但自他上任以来，民主党仍主导了特别选举和中期选举。
然而，这一切并不意味着美国一切安好。特朗普政府的许多试图破坏法治的行动都取得了成功。他释放了攻击国会大厦的“叛乱者”，同时阻止了对瑞妮·古德遇害事件的联邦调查。他清除了忠于宪法的职业检察官，取而代之的是忠于自己的盟友。他的政府已经废除了许多法院命令，并且至少据称侵犯了移民拘留者的权利。此外，还迫使新闻机构和律师事务所做出让步，以避免法律麻烦。更糟糕的是，只要共和党控制参议院，美国司法系统将随着时间的推移越来越倾向于顺从特朗普，因为他的亲信法官占据了越来越多的席位。因此，尽管特朗普政府的意图如预期般恶劣，但美国社会对其民主攻击的抵抗比人们担心的要顽强得多。
A little over a year into his second presidency, President Donald Trump has yet to turn America into a pure, uncut authoritarian nightmare — but not for lack of trying.
Back in January 2025, if you asked a liberal what the new Trump administration would do in the worst case scenario, they would probably have said things like:
• Bring bogus charges against its political enemies, up to and including Democratic lawmakers.
• Strong-arm television networks to de-platform critical comedians.
• Deploy the military to Democratic cities to put down protests against its policies.
• Empower immigration enforcement agents to kill American citizens with total impunity.
The president has already done all of these things.
And yet, on each front, pushback from the judiciary and civil society has either beaten back his assertions of dictatorial power or at least constrained them.
This week brought a particularly vivid illustration of this dynamic in which Trump cooks up some cartoonishly tyrannical plot, only for America’s citizens and constitutional order to force him into retreat.
On Tuesday, Trump’s Justice Department attempted to secure an indictment against six Democratic lawmakers for the crime of exercising their First Amendment rights.
Specifically, the Democrats — including Senators Mark Kelly and Elissa Slotkin — released a video last fall in which they warned those serving in America’s military and intelligence agencies that “threats to our Constitution aren’t just coming from abroad, but from right here at home.”
The lawmakers, all former soldiers or intelligence operatives themselves, reminded their counterparts that they “must refuse illegal orders.” They did not explicitly question the legality of any specific military operation.
But the video came months after Defense Secretary Pete Hegseth allegedly ordered the military to “kill everybody” aboard a Venezuelan fishing boat — a command that was illegal on an array of different levels.
Some might struggle to see what law the Democrats’ message could have broken: How could it be illegal for an elected official to encourage compliance with the constitution or Geneva Convention?
Alas, the president decided that this was not only a crime, but a capital one. On Truth Social, Trump declared the Democrats had committed “SEDITIOUS BEHAVIOR, punishable by DEATH!” and reposted a supporter’s call to “HANG THEM GEORGE WASHINGTON WOULD !!”
In this administration, it seems, the Justice Department’s first loyalty is to the president’s whims rather than America’s laws. So, it was shocking but unsurprising that the US attorney’s office in Washington actually tried to indict the so-called “seditious six” this week.
Yet it was also remarkable that the DOJ failed. Grand juries almost always side with the prosecution, as they hear only the government’s side of a given case. And to approve an indictment, they must merely determine that there is “probable cause” that the accused may have committed a crime.
Yet, the administration could not clear this low bar.
This is part of a broader pattern. The DOJ’s attempts to prosecute some of Trump’s other adversaries, such as former FBI director James Comey and New York Attorney General Letitia James, were also quickly dismissed by judges and grand juries. And although the outcome of the administration’s criminal investigation into Fed chair Jerome Powell remains uncertain, to the extent that it was intended to pressure him into prematurely forfeiting office, it has failed.
This week brought one other, quieter indication of democratic resilience.
In 2025, the Trump administration repeatedly surged federalized National Guard troops to Democratic-run cities in the name of putting down civil unrest — or else, merely deterring protests that the president doesn’t like.
In a June memorandum, Trump authorized the Guard’s deployment to any “locations where protests against” ICE functions “are occurring or are likely to occur based on current threat assessments and planned operations.”
In other words, he asserted the authority to send the military anywhere in the US where a demonstration against his immigration agenda is happening or might soon occur.
This contravened a bedrock norm of American democracy. In the United States, responsibility for upholding civil laws is supposed to lie with civilian officials — not the military — except under the most extraordinary circumstances.
Happily, the administration’s weaponization of the National Guard provoked rebukes from the judiciary, including the conservative-dominated Supreme Court. And on Wednesday, the Washington Post reported that the Trump administration had withdrawn all federalized National Guard troops from U.S. cities.
This basic pattern — of the administration trying but failing to perpetrate a nightmarishly authoritarian act — has repeated itself on many other fronts over the past year.
The Federal Communications Commission tried to bully Disney into taking Jimmy Kimmel off the air for telling jokes it did not like; entertainment unions and Disney+ subscribers threatened to boycott the company if Kimmel were fired, and the network reinstated him.
When Border Patrol agents in Minnesota fired 10 times at the prone body of protestor Alex Pretti, the administration tried to vilify the victim and glorify his killers. But citizens and officials in both parties responded by demanding an investigation into the shooting, and the Justice Department grudgingly obliged.
Meanwhile, the backlash to Pretti’s killing eventually forced the administration to draw down its unprecedented deportation surge in Minnesota, which had been generating routine violations of residents’ civil liberties.
The White House has waged open war on the due process rights of immigrants, with no small amount of success. But one of its most egregious violations of such liberties — the unlawful deportation of Kilmar Abrego Garcia to a foreign prison infamous for torturing its inmates — was rebuffed by federal judges. And the administration eventually repatriated Garcia.
Likewise, the administration tried to deport a Tufts student for authoring an op-ed it disagreed with, only to lose in court.
Perhaps, most importantly, although the president has broadcast an interest in unduly influencing elections in his favor, Democrats have nonetheless dominated special and off-year elections since Trump took office.
None of this is cause for complacency or comfort. A great many of the Trump administration’s attempts to subvert the rule of law have succeeded.
The president has freed insurrectionists who attacked the Capital in his name while barring a federal investigation into Renee Good’s killing. He has purged the Justice Department of career prosecutors loyal to the Constitution and replaced them with allies loyal above all to himself. His administration has nullified many court orders and, at least allegedly, immigrant detainees’ human rights. It has intimidated news organizations and law firms into granting the president concessions to avoid legal trouble.
What’s more: So long as Republicans control the Senate, America’s judiciary is bound to grow more deferential to Trump over time, as his handpicked judges occupy an ever larger share of its seats.
The point then is not that things in America are going well, only that, were the president able to fully execute his will, they would be going incalculably worse.
One year in, this administration’s intentions have proven about as nefarious as predicted. But American society has resisted his assaults on democracy better than one might have feared.
2026-02-13 19:30:00
2026年2月9日，蒙大拿州大瀑布市（Great Falls）的弗里德（Froid）镇居民身穿奥罗兹科柴油（Orozco Diesel）运动衫，以示对罗伯托·奥罗兹科-拉米雷斯（Roberto Orozco-Ramirez）在法庭上接受审讯的支持。在竞选期间，唐纳德·特朗普曾承诺将“大规模驱逐”移民作为其政府的核心政策之一。当时，这一政策似乎很受欢迎，因为民调显示多数人支持驱逐无证移民（尽管并非没有争议）。然而，特朗普的移民执法政策在现实中却导致了美国民众支持率的大幅下降。
在民主党支持率较高的城市如明尼阿波利斯，移民执法问题受到了广泛关注。但农村地区——许多地方都曾大力支持特朗普——同样未能幸免。弗里德这个人口仅195人的小镇就是一个例子。2023年1月，当地一位深受喜爱的机械师罗伯托·奥罗兹科-拉米雷斯被移民官员逮捕，引发了小镇的震动。居民们惊讶地发现，全国范围的移民执法行动不仅限于像明尼阿波利斯和芝加哥这样的大城市，还波及到了这个偏远小镇。
“他们认为特朗普的政策是针对最坏的移民——罪犯、帮派成员等，”蒙大拿自由报（Montana Free Press）的记者诺拉·马比（Nora Mabie）告诉《今日解释》（Today, Explained）节目。“而在这样一个人人相识的小镇，这体现了个人与政治之间的紧张关系，以及当一个社区的投票倾向与个人经历发生冲突时会发生什么。”
《今日解释》主持人诺埃尔·金（Noel King）与马比交谈，了解了奥罗兹科-拉米雷斯的背景、居民对他的逮捕的反应，以及这些反应如何反映出美国人对特朗普移民政策的看法。以下是他们对话的节选（已进行删减和润色）。
这位名叫奥罗兹科的人是一位42岁的父亲，有四个孩子。他已经在蒙大拿州弗里德镇生活了十多年，建立了自己的柴油修理店，为半挂车、校车、拖拉机等提供维修服务，对当地农民来说至关重要。他还是一名小联盟棒球教练，并曾在某段时间非法进入美国。根据法庭文件，他在2009年曾被美国移民与海关执法局（ICE）驱逐出境。之后他返回了美国。2023年1月25日，奥罗兹科-拉米雷斯被边防巡逻队逮捕。投诉文件中提到，执法官员身穿便衣、驾驶无标志车辆，来到他的修理店敲门。他的儿子们说，父亲一开始对这种做法感到怀疑，因为在这个社区，人们通常直接走进店里，而不是敲门。他最终关上了门，据他妻子所说。投诉文件还提到，他离开时曾对执法人员大喊大叫，并在他们开车离开时扔了一块两英尺乘六英尺的木板。这最终导致他被指控威胁联邦官员。但他的家人对此表示异议。
人们对此感到震惊，他们没想到全国范围的移民执法行动会延伸到像弗里德这样偏远的小镇。他们原本认为特朗普的政策只针对最严重的罪犯。而在这样一个社区关系紧密的小镇，这种事件凸显了个人与政治之间的冲突，以及当社区成员的个人经历与他们的政治立场发生碰撞时会发生什么。
在这样一个社区，如果有人敲门，大家都会觉得不对劲，因为这里的人关系非常紧密，通常都是直接走进来。当居民们得知这位男子已被边防巡逻队拘留时，他们的反应如何？对于镇上的居民来说，这是他们第一次了解到罗伯托的移民身份。很多人表示他们并不知道他在这里非法居住。他们只是把他当作邻居和朋友。当他们得知他曾被驱逐出境，这确实引发了一些复杂的思考。有些人说：“规则就是规则，他应该遵守规则，走合法途径获得公民身份。”但他的儿子们则希望更多人理解，获得美国合法公民身份并不容易。
显然，人们非常重视这位邻居。他们对移民执法在小镇上发生有什么看法？有些人告诉我，这里的人确实支持特朗普，但并不是因为他的移民政策。他们更关心的是经济问题，以及影响农业或石油产业的议题。我在镇上的酒吧里和一位居民交谈，他说他非常支持特朗普在边境的政策，喜欢他的做法。但当我问他如何看待罗伯托时，他说罗伯托应该有机会合法获得公民身份。
这让人看到这些相互冲突的观点。人们并不是单纯支持或反对移民政策，而是大多持混合态度，这也是这个故事引人注目的地方。让我印象最深的是，有14位社区成员，占小镇人口的相当大比例，驱车七小时前往罗伯托的首次听证会。而那次听证会只持续了八分钟。之后，我看到他们围在一起，互相拥抱，有些人甚至流泪，然后他们又坐上车，返回家中。
我们在阅读你的报道时，有评论说，人们在理论上支持特朗普的移民政策，但当他们意识到这些政策实际上如何影响他们和邻居时，就感到震惊。你是否也看到了这种现象？
在弗里德，人们并不喜欢谈论政治和分歧。我甚至和镇长谈过，他说我们这个社区太紧密了，如果你说了不当的话或惹恼了某人，你的生意可能会受到严重影响。在这个镇上，你需要依靠邻居们。因此，我并没有听到人们说他们会改变投票意向。当我与一些社区成员交谈时，他们表示不确定。但有些人提到，最近几天，邻居们院子里的特朗普支持标语已经不见了。奥罗兹科的儿子们还说，他们在Facebook上看到至少有一条评论表示，他们为曾投票支持特朗普感到抱歉。
你认为这个小镇的故事反映了我们当前政治局势的什么？我认为这表明特朗普政府实施的移民政策已经影响到了几乎每个人。如果这些政策能波及到这个偏远的农村小镇，而这位男子在这里已经生活了十年，没有造成任何问题，同时引发了社区和周边社区的强烈反应，这说明这种执法行动的普遍性以及其带来的连锁反应。如果罗伯托最终被长期拘留或被驱逐出境，他的家庭将失去父亲，人们将失去朋友和邻居，而许多农民也将因为失去这位机械师而遭受损失。镇长甚至表示，当地经济将受到重大打击。这个地区本身商业不多。许多美国人正在努力应对特朗普移民政策带来的真正困难的后果。我认为一个大问题是，这些政策将如何影响他们的投票行为，以及是否会改变他们的政治立场。
On the campaign trail, Donald Trump vowed the “mass deportation” of immigrants would be a centerpiece of his administration. At the time, it looked like a winner: Polls found majority support for mass deportation of undocumented immigrants (though not without complications).But the reality of Trump’s immigration enforcement policies has caused Americans’ support to tank.
The fight over immigration enforcement in Democratic strongholds like Minneapolis has gotten the most attention. But rural areas — many of which voted overwhelmingly for Trump — are not immune. One of them is the tiny town of Froid, Montana.
In January, the arrest of a beloved mechanic, Roberto Orozco-Ramirez, by immigration agents roiled the 195-person town. Residents were surprised to see the national immigration crackdown extend beyond big cities like Minneapolis and Chicago.
“They believed Trump’s policy to be about arresting the worst of the worst — criminals, gang members, things like that,” Nora Mabie, a reporter at the Montana Free Press, told Today, Explained. “And in this town where everybody knows everybody, it’s an example of the tension between the personal and the political and what happens to a community when they vote one way, but then a personal experience forces them to confront their beliefs.”
Today, Explained co-host Noel King spoke with Mabie to learn more about Orozco-Ramirez, how residents of Froid have responded to his arrest, and what it tells us about how Americans are thinking about Trump’s immigration policies. Below is an excerpt of their conversation, edited for length and clarity. There’s much more in the full episode, including interviews with Froid residents, so listen to Today, Explained wherever you get podcasts, including Apple Podcasts, Pandora, and Spotify.
Tell me about this man, Mr. Orozco.
He is a 42-year-old father of four. He has lived in Froid, Montana, for more than a decade, long enough to build his own diesel shop where he fixes semi trucks, school buses, tractors, and really provides an invaluable service to farmers in the area. He’s also a Little League coach and he came to the US at some point illegally. Court documents say that in 2009 he was removed by ICE and deported. At some point he came back.
On January 25, Orozco-Ramirez was arrested by Border Patrol. In the complaint it documents that officers, wearing plain clothes and in unmarked vehicles, came to his diesel shop and knocked on the door. Later when I was talking with his sons, they said their dad was immediately suspicious of this because people never knock on the door. This is a place where people just walk in. So he was suspicious and he ended up kind of closing the door on them. That’s what his wife says.
The complaint also alleges he yelled at them on their way out. And as they were in their car driving away, it says he threw a two-by-six piece of lumber in their direction. And that’s what led ultimately to a charge of threatening a federal officer. That’s something his family disputes.
People were shocked. They were surprised to see the national immigration crackdown extend beyond big cities like Minneapolis and Chicago and reach this tiny corner of the state. They said they believed Trump’s policy to be about arresting the worst of the worst — criminals, gang members, things like that. And in this town where everybody knows everybody, it’s an example of the tension between the personal and the political and what happens to a community when they vote one way, but then a personal experience forces them to confront their beliefs.
So this is a town where you know something’s up if someone knocks at your door because it’s so tight-knit that normally people just walk on in. How do people in this tiny, tight-knit town react when they learn this man is now in the custody of Border Patrol?
For people in town, this is the first time they are learning about Roberto’s status. A lot of people told me they didn’t know he was here illegally. You know, they just thought of him as their friend and their neighbor. And yeah, when they hear that he was deported once, it does bring up some complicated thoughts.
Some people are saying, “Rules are rules, you know, he should have followed the rules and done the path to obtain citizenship.” But that’s [where] his sons, you know, wish more people would understand that gaining legal citizenship in the US is not that easy.
Obviously people feel very strongly about this man, their neighbor. How do they feel about immigration enforcement happening in their little town?
Some people told me that people really support Trump here, but that they don’t necessarily support him because of his immigration policies. The issues that are important to people here are the economy and things that affect farming or even oil.
I did talk to this man in the bar. He said he really likes what Trump’s doing at the border, and he likes that policy. But when I asked him about what he thought of Roberto, he said he thought Roberto should have the opportunity to pursue legal citizenship.
It’s interesting to see these really conflicting beliefs on display. People are not pro or con on this issue, they’re mostly mixed, which is another thing that makes this story so interesting. You know what really struck me is 14 of these community members, which is a pretty significant percentage of the town’s population, drove seven hours one way to Roberto’s initial hearing. And that hearing only lasted eight minutes. And then I just watched them as they huddled after the hearing, they hugged each other, some people cried, and then they got back in their cars and drove home.
We were reading your story and someone in the comments said something along the lines of people were happy to vote for the Trump immigration policy in theory, but then they realized how it actually affects them and affects their neighbors. Does that feel like what you saw?
In Froid, people do not really like talking about politics and division. I even spoke with the mayor, who said we’re too close-knit for that — if you say the wrong thing or upset someone, your business could totally go under. You need your neighbors in this town. So I didn’t quite hear people saying that they would change their vote. And when I talked to some community members, they said they weren’t sure.
But then some people said Trump signs that were displayed in people’s yards in the neighborhood have come down in the last few days. And then the Orozco boys also said they saw on Facebook at least one comment from someone saying they were sorry that they voted for Trump.
What do you think this story from this small town tells us about the political moment we’re in more broadly?
I think it’s shown that these immigration policies that the Trump administration is implementing have affected almost everybody. If they are stretching to this tiny rural town where this man has lived for 10 years without problem, and really riling up this community and neighboring communities, I think it just shows the ubiquity of this crackdown and also the ripple effects from it.
If Roberto ends up detained for a long time or if he is deported, a family will lose their father, people will lose their friend and neighbor, [and] a lot of farmers will suffer in this area by losing their mechanic. And the mayor even said the local economy will take a big hit. There’s not a lot of businesses in the area.
Many people across the country are grappling with the really difficult consequences of Trump’s immigration policies. And I think the big question is, how will this change, or will this change, how they vote, and will it change their politics in the future?
2026-02-13 07:40:00
这并不是你下个月的办公室。在X平台上一篇名为《Something Big Is Happening》的病毒式文章中，Matt Shumer写道，世界正在经历一个类似于人工智能早期发展的时刻。OthersideAI的创始人兼首席执行官Shumer认为，人工智能已经从有用的助手转变为具有普遍认知能力的替代品。更重要的是，AI现在正在帮助构建更强大的自身版本。很快，能够媲美人类大多数专业技能的系统可能会出现。虽然专家们知道变革即将到来，但普通大众可能即将措手不及。用疫情时代的比喻来说，Tom Hanks即将生病。
在Shumer的文章发表后，Anthropic安全团队负责人Mrinank Sharma辞职，他在离职信中警告称，“世界正面临危险”来自“相互关联的危机”，并暗示公司“不断面临压力，要放弃最重要的事情”以追求3500亿美元的估值。因此，一些人开始感到焦虑。更准确地说，那些已经对AI非常担忧的人现在更加焦虑了。
当然，我们不能忽视AI模型可能很快就会满足各种所谓的弱AGI（通用人工智能）定义，至少在某种程度上。许多技术专家和预测市场都持这种观点。不过，与其关注技术进步——尽管我确信生成式AI是一种强大的通用技术——不如从经济学的角度来看待一些基本的瓶颈和限制。从演示到实际部署的漫长道路。从“AI模型令人印象深刻，甚至比你意识到的还要厉害”到“一切都将立即改变”的跳跃，需要忽略经济系统如何真正吸收新技术。电力普及花了数十年才重新设计工厂。互联网并未一夜之间改变零售业。目前，美国不到五分之一的企业正在采用AI。在大型、受监管且风险规避的机构中部署AI，需要大量配套投资，包括数据基础设施、流程再造、合规框架和员工再培训。（经济学家称之为“生产率J曲线”）事实上，早期的投入可能会在可见成果出现之前抑制产出的测量值。富有并不总是意味着更忙碌。
让我们承认乐观者的观点——我当然也认为自己相当乐观——他们假设AI能力迅速提升。但产出并不会立刻激增。历史上，富裕社会往往选择更多的休闲时间，比如提前退休或缩短工作周，而不是在办公室或工厂里多花时间。经济学家Dietrich Vollrath指出，如果家庭因生产力提高而减少劳动供给，那么更高的生产力并不一定意味着更快的经济增长。福利可能大幅上升，但GDP增长却相对温和。
最慢的行业决定了整体增长速度。即使AI使某些服务变得非常便宜，需求也不会无限增长。支出会转向那些难以被自动化的行业，如医疗、教育和面对面体验，这些行业的产出更紧密地与人类时间相关。（这就是著名的“鲍莫尔效应”或“成本病”）随着整体工资上涨，劳动密集型且生产力增长缓慢的行业将占据更大的收入份额。结果是：即使AI取得显著进展，整体生产力增长可能也仅是适度的。
在由许多互补部分组成的系统中，经济学家Charles Jones指出，最狭窄的瓶颈决定了整体的流动。AI可以加速编码、起草和研究，但若能源基础设施、物理资本、监管批准或人类决策仍以普通速度推进，这些将成为限制整个经济快速增长的瓶颈。
经济系统是适应性强、复杂且美妙的系统。它们创造了体现并积累复杂信息的物理产品——经济学家Cesar Hidalgo称之为“想象力的结晶”。当它们发生变化时，它们会通过渐进的重组和重新配置来调整，而不是突然崩溃或立即起飞。因此，这应该是你的基本预期。
当然，一定程度的紧迫感可能是必要的。（Shumer建议现在就拥抱最强大的AI工具，并将其融入日常工作中，这似乎是有道理的。）但将AI与2020年初的疫情进行令人恐慌的类比则不太合适。本文最初发表于Pethokoukis的通讯《Faster, Please!》。
In a viral essay on X, “Something Big Is Happening,” Matt Shumer writes that the world is living through a moment similar to early Covid for artificial intelligence. The founder and CEO of OthersideAI argues that AI has crossed from useful assistant to general cognitive substitute. What’s more, AI is now helping build better versions of itself. Systems rivaling most human expertise could arrive soon.
While experts know transformative change is coming fast, normies are about to be blindsided. To stick with the pandemic-era metaphor, Tom Hanks is about to get sick.
Between Shumer’s essay and the resignation of Mrinank Sharma — he led Anthropic’s safety team and vague-posted quite the farewell letter warning that “the world is in peril” from “interconnected crises,” while hinting that the company “constantly face[s] pressures to set aside what matters most” even as it chases a $350 billion valuation — well…some people are starting to wig out. Or, more precisely, the folks already super-worried about AI are now super-worrying even harder.
Look, is it possible that AI models will soon indisputably meet various so-called weak AGI definitions, at minimum? Plenty of technologists, not to mention prediction markets, suggest it is. (As a reality check, though, I keep front of mind Google DeepMind CEO Demis Hassabis’s statement that we still need one or two AlphaGo-level technological breakthroughs to reach AGI.)
But rather than technological advances — and I have high confidence generative AI is a powerful general-purpose technology — let’s instead talk about some basic bottlenecks and constraints from the world of economics rather than computer science.
The long road from demo to deployment. The leap from “AI models are impressive, even more than you realize” to “everything changes imminently” requires ignoring how economies actually absorb new technologies. Electrification took decades to redesign factories around. The internet didn’t change retail overnight. AI adoption currently covers fewer than one in five US business establishments. Deploying it across large, regulated, risk-averse institutions demands heavy complementary investment in data infrastructure, process redesign, compliance frameworks, and worker retraining. (Economists term this the productivity J-curve.) Indeed, early-stage spending can actually depress measured output before visible gains arrive.
Richer doesn’t always mean busier. Let’s grant the optimists — and I certainly consider myself pretty darn optimistic — their assumption about fast-advancing AI capability. Output still doesn’t explode on a dime. Richer societies historically choose more leisure — earlier retirements, short workweeks — not more time at the office or factory floor. Economist Dietrich Vollrath has pointed out that higher productivity doesn’t mechanically translate into faster growth if households respond by supplying less labor. Welfare might rise substantially while headline GDP growth stays relatively modest.
The slowest sector sets the speed limit. Even if AI makes some services far cheaper, demand does not expand without limit. Spending shifts toward sectors that resist automation — health care, education, in-person experiences — where output is tied more tightly to human time. (This is the famous “Baumol effect” or “cost disease.”) As wages rise economy-wide, labor-intensive sectors with weak productivity growth claim a larger share of income. The result: Even spectacular AI gains may yield only moderate growth in overall productivity.
The economy’s narrowest pipe. In a system built from many complementary pieces, explains economist Charles Jones, the narrowest pipe determines the flow. AI can accelerate coding, drafting, and research all it wants. But if energy infrastructure, physical capital, regulatory approval, or human decision-making move at ordinary speeds, those become the binding constraints that limit how fast the whole economy can grow.
Economies are adaptive, complex, wonderful systems. They create the physical objects that embody and accumulate complex information — what economist Cesar Hidalgo elegantly calls “crystals of imagination.” And when they change, they adjust through gradual reorganization and reallocation, not through sudden collapse or instant takeoff. I mean, that should be your baseline scenario.
Now, a degree of urgency may be warranted. (Shumer’s advice to embrace the most capable AI tools now and weave them into your daily work seems prudent.) Panic-inducing analogies to early 2020 probably are not.
This piece originally appeared in Pethokoukis’s newsletter “Faster, Please!”