2026-09-13T16:00:29.763Z / 美国有线电视新闻网（CNN）
本文原版刊载于CNN的《重要事项》新闻简报。如需免费订阅并将其发送至您的邮箱，请点击此处注册。
最新的人工智能预警似乎旨在颠覆认知、敦促所有人采取行动。
人工智能竞赛的领军人物、Anthropic公司首席执行官达里奥·阿莫代伊（Dario Amodei）上周六在其个人网站发布的一篇文章中呼吁有意识地放缓人工智能开发进度。他写道，人工智能的发展速度越来越快，可能会超出人类的管控能力。
“如果不加约束，人工智能可能会超出我们理解和控制这些系统的能力，因此即便要研发，也必须极其谨慎。”阿莫代伊写道。
就在这一呼吁发出几天前，阿莫代伊公司的一名前工程师发出了更为极端的警告，称人工智能可能会导致人类灭绝。
“研发人工智能的人真心认为，到本世纪末，人工智能可能会让我们全军覆没。”雅各布·考克森（Jacob Coxon）说道。他曾在OpenAI和Anthropic两家公司任职，后来在社交媒体上发布了引发热议的警告。
正如考克森向CNN主持人安德森·库珀解释的那样，没人知道人工智能究竟会以何种方式终结人类。
他们的论点是，人工智能的智能水平一直在提升，很快就能实现自我迭代，超越人类的智慧，进而引发……一系列灾难。比如劫持核武器？释放毁灭性病毒？让世界陷入黑暗？打造一个让人找不到工作、毫无成就感的糟糕社会？随便哪种都有可能。
或许能让人稍感安心的是，尽管美国国防部正依赖人工智能，但美国国防部长皮特·赫格斯西特（Pete Hegseth）曾表示：“人工智能不会做出致命决策。”除此之外，法律并未明确限制人工智能在所谓“杀伤链”中的使用范围。
考克森的言论得到了一些公益组织和议员的呼应，他们希望采纳那些呼吁政府出台监管框架、限制人工智能开发方式的人工智能企业的建议。
Anthropic和OpenAI等公司已公开呼吁政府对其进行监管，以放缓当前的人工智能主导权竞赛——这与它们冲刺首次公开募股（IPO）的节奏不谋而合。
科技记者、CNN特约撰稿人雅各布·沃德（Jacob Ward）周四在接受CNN主持人卡西·亨特采访时表示，我们正在目睹“一个规模很小却实力强大的企业集团联手，朝着他们口中的悬崖狂奔，却似乎停不下来”。
他认为，不应忽视这些正在赚得盆满钵满的人发出的警告。
“他们在说：‘嘿，我们正在研发这项技术，这很令人兴奋，但它可能不仅会让你丢掉工作，还会夺走你的生命。’我认为我们现在开始重视这件事至关重要。”沃德说道。
呼吁政府采取行动的人士中包括参议员伯尼·桑德斯（Bernie Sanders），他认真对待这些警告。
“几乎每天都有令人恐惧的新报道，讲述科技巨头正在失去对其研发技术的控制，可能带来灾难性后果。”桑德斯在宣布拟立法禁止超级人工智能时说道，“社会放任他们继续推进、让这些产品变得更先进，这是不负责任的。”
桑德斯还就自己的提案接受了CNN主持人凯特兰·柯林斯的采访。
科技投资人、特朗普总统科学技术顾问委员会联合主席戴维·萨克斯（David Sacks）则持相反观点，主张采取基于开放市场的人工智能发展路径。
萨克斯周四在福克斯新闻节目中露面时，质疑了考克森的动机，并暗示其警告是“末日恐慌论者”阴谋的一部分，或许背后有中国资助，目的是阻碍美国的人工智能创新。
周六，唐纳德·特朗普总统呼应了这一论调，他在爱尔兰的一场新闻发布会上表示，那些担忧人工智能最终影响的人“提出的事情根本不会发生”。
当被问及该行业是否需要监管时，特朗普重申了此前关于与中国竞争的表态。
“我们在人工智能领域领先于中国。”他说，“我们是世界上技术最先进的国家，坦率地说，我希望保持这一地位，因为谁在人工智能领域获胜，谁就掌握了主动权。我们可以设置防护栏。”
萨克斯是那些认为美国企业必须继续创新，不能让中国取得人工智能霸权的人士之一。
“中国没有像我们国家这样庞大的末日恐慌团体影响力运作。”萨克斯在接受福克斯新闻主持人布雷特·拜尔采访时说道，“但我们美国国内，有一场声势浩大的反对人工智能的媒体宣传活动，正在让公众对人工智能产生抵触情绪。”
他还提到中国：“他们肯定乐见我们在这里停止人工智能的研发。”
在萨克斯发布到X平台的“All-In”播客视频中，他辩称人工智能企业呼吁监管，实则是为了有效掌控人工智能。他将此称为“前沿市场双头垄断”。
他认为，政府官员根本不知道该如何监管人工智能。
“所以他们会求助于顶尖实验室，说：‘嘿，我们该如何监管它？’当然，Anthropic会给出所有答案。”他说道，暗示人工智能失控的警告是一场争夺市场主导权的把戏。
萨克斯认为，与其说人工智能有终结人类的潜力，不如说将人工智能的权力集中在少数人手中才是真正的危险。他在另一篇X平台帖子中表示：
“保持创新的去中心化和可及性，是避免陷入不安全、反乌托邦未来的关键——到那时，先进的人工智能能力将被集中掌控在少数人手中。”
这种想法与特朗普的立场一致：在华盛顿未能出台相关法规的情况下，特朗普试图阻止加利福尼亚等州自行制定人工智能监管规则。
去年，国会搁置了禁止各州制定相关法规的提案，特朗普随后签署了一项行政命令，试图达成同样的目标。
正如CNN记者哈达斯·戈尔德（Hadas Gold）在其近期报道《有点像新冠疫情初期：人工智能监管的混乱争夺》中所解释的那样，特朗普政府迄今为止更多地将人工智能监管视为与中国在该技术领域的竞赛的一部分，而非保护人类的举措。
例如，该政府曾阻止部分英伟达芯片出口到中国，或暂时推迟产品发布以评估其对国家安全的影响。
对于那些认为人工智能存在生存危机的人来说，这一切可能会让人联想到冷战时期的核军备竞赛——当时各国最终搁置分歧，在20世纪60年代及之后签署了一系列条约，包括《不扩散核武器条约》、核试验禁令、战略武器限制条约等。
但很难想象特朗普政府——或是中国——会参与任何全球条约谈判，尽管欧洲曾尝试制定人工智能监管条约。
核武器构成的威胁仍然是美国外交政策的核心关切，也是特朗普发动对伊朗战争的主要理由之一。
尽管关于人工智能最极端的警告涉及人类灭绝，但作为一个议题，它最有力的激励点在于选民对人工智能侵入自身生活的担忧。
美国民众对遍布各地、支撑人工智能集群的数据中心感到不安，科技巨头计划在各地建设这类数据中心，这会影响水质、推高能源账单，并占用农田。
这种类似民粹主义抗议的反对大型数据中心建设的浪潮，是今夏的政治热点，跨越党派界限，团结了左右两翼民众，就连特朗普这类只看到商业利益的政策制定者也感到困惑。
美国选民通常不会着眼长远。
他们会担忧气候变化，但投票时却更关注油价。他们对40万亿美元的国家债务感到震惊，但往往会投票给减税并增加开支的议员。而在人工智能问题上，他们似乎更在意数据中心带来的困扰，而非潜在的生存威胁。
但也有可能，民众对数据中心的反感，反映出他们将数据中心视为生成式人工智能最直观的物理体现。科技公司承诺这项技术将改变世界，一些专家担心它可能会终结人类——但许多普通美国人只是觉得它笨拙又令人反感，他们并不希望人工智能侵入自己生活的诸多领域。
AI and the end of humanity? OK, ‘doomer’ says one Trump official
2026-09-13T16:00:29.763Z / CNN
A version of this story appeared in CNN’s What Matters newsletter. To get it in your inbox, sign up for free here.
The latest warnings about AI seem intended to boggle minds and shock everyone into action.
Anthropic CEO Dario Amodei, a leader in the AI race, on Saturday called for a conscious slowdown in AI development in an essay posted to his website. AI is developing faster and faster and could outrun humans’ ability to manage it, he wrote.
“Left unchecked, it could outrun our ability to understand and control these systems, and so must be pursued very carefully, if at all,” Amodei wrote.
The call for action comes just days after a former engineer at Amodei’s company went even further, saying AI could bring about the end of humanity.
“The people building AI earnestly believe that it could kill us all by the end of the decade,” said Jacob Coxon, who had worked at both OpenAI and Anthropic before he posted his viral warning on social media.
Nobody knows how exactly AI would end humanity, as Coxon explained to CNN’s Anderson Cooper.
The argument is that AI is getting smarter all the time and will soon make itself smarter on its own, be able to outwit humans and unleash … something. Hack nuclear weapons? Let loose a devastating virus? Plunge the world into darkness? Create a lame society in which unfulfilled people can’t find work? Take your pick.
It should be comforting, maybe, that while the Pentagon is relying on artificial intelligence, Secretary of Defense Pete Hegseth has said, “AI is not making lethal decisions.” Beyond that, the law does not place explicit limits on where AI can be used in the so-called kill chain.
Coxon’s message was amplified by public interest groups and lawmakers looking to heed the advice of the AI companies that want the government to impose a framework and limits on how AI is developed.
Companies such as Anthropic and OpenAI have publicly called for the government to regulate them so it can slow down their current race for AI domination, which coincides with their race to initial public offerings.
What we’re seeing is “a very small and powerful group of companies, linked arms together, running toward what they say is a cliff they cannot seem to stop themselves from running off,” the tech journalist and CNN contributor Jacob Ward told CNN’s Kasie Hunt on Thursday.
He argued that calls from these people who are in the process of making a lot of money should not be ignored.
“They are saying, ‘Hey, we’re building this thing, and it’s exciting, but it might not just cost you your job, but your life.’ I think it is important that we start paying attention,” Ward said.
Among those calling for government action is Sen. Bernie Sanders, who takes the warnings seriously.
“Nearly every day, there is a frightening new story about how Big Tech companies are losing control of the technology they are developing, with potentially cataclysmic results,” Sanders said in announcing proposed legislation to ban artificial superintelligence. “It is irresponsible for society to allow them to move forward and make these products even more advanced.”
Sanders also talked about his proposal with CNN’s Kaitlan Collins.
David Sacks, a tech investor who is also co-chair of President Trump’s Council of Advisors on Science and Technology, has pushed in the other direction, toward an open market-based approach to AI.
Appearing on Fox News Thursday, Sacks questioned Coxon’s motivations and seeded the idea that his warning was part of a conspiracy by “doomers,” maybe financed by China, to thwart AI innovation in the US.
On Saturday, President Donald Trump echoed that sentiment, saying during a news briefing in Ireland that people worried about the ultimate impacts of AI “are bringing up things that won’t happen.”
Asked whether the industry needs to be regulated, Trump said repeated his earlier statement about competing with China.
“We’re leading China in AI,” he said. “We’re the most sophisticated country in the world, and frankly, I want to keep it that way, because whoever wins AI wins. And we can put guard rails.”
Sacks is among those who argued American companies must continue to innovate rather than allow China to achieve AI supremacy.
“They don’t have this giant influence operation of these doomer groups in China,” Sacks told Fox News’ Bret Baier. “What we have here in the US, though, is we have this huge media campaign against AI and it is turning the public against AI.”
Sacks added of China: “They would love for us to stop AI development over here.”
In a video from his “All-In” podcast, which Sacks posted to his account on X, he argued the AI companies want regulation so that they can effectively control AI. He called it a “frontier market duopoly.”
Government officials aren’t going to have any idea how to regulate AI, he argued.
“So they’re going to turn to the leading labs to say, ‘Hey, how should we regulate it?’ And of course, Anthropic is going to have all the answers on that,” he said, suggesting warnings of AI run amok are part of a play for market domination.
Rather than push the idea that AI has the potential to end humanity, Sacks sees a conspiracy to place power over AI in a few select hands. That, he said in another X post, would be the true danger.
“Keeping innovation decentralized and accessible is the key to avoiding an unsafe and dystopian future where advanced AI capabilities are centralized and controlled by only a few hands,” he said.
That kind of mindset is in line with Trump’s efforts to stop states like California from imposing their own AI regulations in the absence of movement from Washington.
When efforts to ban state regulations stalled in Congress last year, Trump signed an executive action trying to do the same thing.
The Trump administration has so far viewed AI regulation more as part of a race with China over the technology than as an effort to protect humanity, as CNN’s Hadas Gold explained in her recent story, ‘It feels a little like early Covid’: The messy scramble to regulate AI.
The administration blocked export of some Nvidia chips to China, for instance, or temporarily delayed release of products to assess their impact on national security.
To those who see existential danger in AI, all of this may echo the nuclear arms race of the Cold War, when countries ultimately set aside differences to impose a series of treaties in the 1960s and beyond for Nuclear Non-proliferation, test bans, strategic arms limitations, and more.
But it’s hard to imagine the Trump administration — or China — getting involved in any worldwide treaty effort, although Europe has tried to go the route of a treaty for AI regulations.
The threat posed by nuclear weapons still animates US foreign policy and is a main reason cited by Trump for his war on Iran.
While the most dire warnings about AI have to do with the end of humanity, it is most potent as an issue motivating voters worried about AI encroaching in their backyards.
Americans are unnerved at unsightly data centers that power the hive mind of AI and which tech giants would have dot the landscape, impacting water quality, driving up energy bills, and swallowing up farmland.
The populist, pitchfork-style opposition to the massive construction projects has been the political story of the summer, crossing party lines and uniting people on the left and right, even as it confuses policymakers like Trump, who see dollar signs.
American voters cannot typically be accused of looking long-term.
They express concern about climate change, but seem to vote based on gas prices. They shudder at the $40 trillion national debt, but often vote for lawmakers who cut taxes and increase spending. And with AI, they appear to be far more motivated by concerns about data centers rather than the potential existential threat.
But it’s also possible that distaste for data centers reflects their role as the most obvious physical manifestation of generative AI. Tech companies promise the technology will change the world, and some experts worry it may end humanity — but many ordinary Americans just find it clunky and off-putting, and they didn’t ask for it to invade so many areas of their lives.
发表回复