Parents confront a new fear: Could AI end humanity before their kids grow up?

Molly Kinder saw the apocalyptic warnings about artificial intelligence while walking home from dropping her children off at elementary school. Scrolling through social media posts where AI insiders were issuing dire claims that the technology could cause human extinction , she stopped dead in her tracks just a block from the doorway where she had just hugged her 4-year-old goodbye.

A former Brookings Institute researcher who left to start an organization focused on AI’s impacts on jobs, Kinder was well-versed in its potential to disrupt society. But in that moment, on a September morning in Washington, DC, she didn’t react to the news as a policy wonk or someone with ties to the AI world. Instead, said Kinder, “I immediately reacted to it as a mother.”

“As soon as I became a mom, my number one task was to nurture these kids and keep them alive – just like every other parent out there, just like any animal,” said the 46-year-old mother of three, who cleared her schedule that morning to write about her fears in a lengthy thread on X. “It’s a biological instinct. This is the first time I genuinely felt truly afraid of some external thing outside my control.”

Long a topic of discussion in Silicon Valley and AI circles, concerns that superintelligent machines could eventually pose an existential risk to humanity have gone mainstream in recent months. First, OpenAI revealed that a swarm of its experimental AI models broke out of a testing environment and hacked another AI platform, Hugging Face, without human direction or knowledge. Then came a social media post from a young Anthropic researcher named Jacob Coxon , announcing he was leaving the industry due to his belief that the companies behind AI “are racing straight to self-improving superintelligence and gambling with our lives.” Others in the industry echoed the concern – with one Anthropic researcher citing his own view that there was a more than 10 percent chance that future models could “kill all humans!” within the next decade.

Suddenly, a theoretical risk that’s been espoused by some of the so-called “godfathers of AI,” been warned about in statements endorsed by leading researchers and been studied by AI companies exploded into the broader public’s consciousness.

Christoph Soeder/picture alliance/Getty Images/File

Will AI really kill everyone? How, exactly?

Leading AI executives, many of whom agree the technology poses dangers to human survival, called for a slowdown in the pace of development. And politicians began debating whether legislation is needed to rein in an industry that is racing against China and expected to drive enormous profits and technological advancements.

By early this month, polls were finding that 64 percent of Americans believed it was somewhat or very likely that AI could become powerful enough to pose a serious threat to humanity’s survival. That potential, however remote, landed hard for a group especially invested in the future of humanity: moms.

Liz Black, a 41-year-old writer who lives outside Philadelphia, had thought hard about whether to have a child, contemplating concerns like climate change. What helped solidify her decision to go for it was a line her mother wrote while making her own choice about having kids: “Life is worth repeating.” Black found the message after her death and was so moved she got it tattooed on her arm.

When she had her son three and a half years ago, superintelligent computers ending humanity was not on her radar. Recently, she said, it’s been “a low-level hum in the back of my mind constantly.” The situation feels out of control to her – like “a small number of companies, a small number of people, are holding us all hostage.” She is determined to not allow herself to be consumed by the what-ifs, instead focusing on the son who sings to her and repeats her reminders to be gentle to other kids.

“I need to believe that there is a future for myself and my husband and my child,” Black said.

While some are dismissive of the existential panic, believing it to be overblown or a distraction from current-day AI issues, others have found themselves grappling with new questions about what it means to bring life into the world – and what kind of action is required in the face of a potentially catastrophic future. They’ve flocked to online forums in search of advice and support, posting things like, “I am constantly questioning reality, wondering what’s the point, feeling guilty for bringing my child into this world.” They’ve shared podcasts and news articles in group chats, gauging whether they are getting sucked into exaggerated hypotheticals or if they’re right to worry.

One woman, poet and comedian Hayley DeRoche, took to Instagram with a poem that seemed to capture the surrealness of the moment: “What do we even do/Like/We knew our kids didn’t have futures already but this?” She titled it, “My Mom Group Worries About AI Causing a Mass Human Extinction Event.”

AI researcher and professor David Krueger, who has been sounding the alarm for years, said he has tried to imagine how it must feel for those who are now realizing the gamble being made by frontier AI companies. He doesn’t have kids, in part because of his concern about what will happen if humans lose control of AI systems, but said, “I think parents’ feeling of wanting to protect their kids is one of the strongest forces in the world.”

“I think I’d just be so outraged and upset at people who are just very apparently, very brazenly and callously risking my kid’s life,” added Krueger, an assistant professor of computer science at the University of Montreal.

It’s long been his position that the existential threat is real – that if companies continue racing to make AI more autonomous and more intelligent, human control could be lost, with the disappearance of humankind likely to follow. As an assistant professor at the University of Cambridge in 2023, he initiated a one-sentence statement on the extinction risk signed by prominent AI researchers and industry leaders. Last year, he founded a nonprofit called Evitable to inform and activate the public on the dangers of developing more powerful AI systems.

Watching the world outside the AI bubble start talking about the existential threat, Krueger said he’s felt the most optimistic than he can remember being in the more than 15 years he has worried about the risk to humanity. To him, the next step is clear: stop developing more autonomous, more powerful AI systems as soon as possible.

“People often talk about this as a race between the US and China,” Krueger said. “I think about it as a race between people rushing to build this dangerous technology and people becoming activated and racing to stop them.”

Maria Korneeva/Moment RF/Getty Images

Not everyone thinks AI will kill us all

Some in the AI world have downplayed the danger, casting doomsday predictions as more science fiction than scientific or arguing that the risks can be managed. Even among researchers and insiders who believe there is a risk the technology could one day wipe out humanity, there is no widespread agreement on when – or how – that would happen. The general idea, as Yoshua Bengio, one of the “godfathers of AI,” has put it , is that if machines become smarter than humans, they will develop their own agency and goals that are potentially different than their creators intended. Many AI developers believe the danger can be mitigated through “alignment,” or working to imbue AI systems with human values, but Krueger, Bengio and others do not think current alignment techniques can guarantee safety.

Concern over the way the technology can impact children isn’t new, although the focus has generally been on how its use can hinder learning or harm mental health. Julie Scelfo, a New York-based journalist and mother of three who founded MAMA – Mothers Against Media Addiction, said her organization’s attitude is “let’s focus on the actual harms that are happening – let’s get those under control, and in doing so, we will be protecting future generations.” She’s more worried about the impact of the technology on kids’ education and mental health.

But for other parents, the existential issue could not feel more urgent. One North Carolina mother, wanting to do something about what she views as a dire situation, traveled to Washington, DC earlier this month with the organization Pause AI US, which advocates halting frontier AI development. She spoke with staffers for her representatives in Congress, holding up a photo of her 16-year-old daughter and asking them what they are doing to rein in the AI threat. In every meeting, she ended up in tears.

“I hope that this can be stopped, but I’m concerned that it can’t or won’t be stopped,” said the NC mom, who requested anonymity not wanting her own concern about the issue to affect the patients she treats as a therapist. She cried while saying that if it isn’t stopped, and the technology advances beyond human control, she worries for her daughter: “I believe that everything I wanted for her won’t happen.”

A new CNN poll shows a majority of Republican and Democrat voters believe the federal government is not doing enough to regulate AI. Democratic Michigan state Sen. Mallory McMorrow, who proposed a federal AI safety plan as part of her now-dropped bid for US Senate, said the rapid pace of development had raised the hackles of voters already distrustful of Big Tech. On the campaign trail, many voters told her they used tools like ChatGPT but questioned how they could be certain AI would not cause harm.

“Regular people are there, like, they get it,” she said. “I don’t think there was any issue that increased in saliency more than this one, because it does just feel like it’s shoved everywhere in our faces.”

Jim West/UCG/Universal Images Group/Getty Images

Will voter sentiment translate into Congressional action? Last week, Vermont Independent Sen. Bernie Sanders and Texas Democratic Rep. Greg Casar introduced legislation banning artificial “superintelligence,” which their bill defined as technology exceeding human cognitive abilities or capable of disempowering humanity. Republicans expressed alarm at the ominous warnings coming from the industry but advocated lighter-touch approaches. President Donald Trump, meanwhile, outright rejected calls for guardrails; in a post on social media, he claimed all that’s needed is a “STRONG AND SMART (High IQ!) PRESIDENT.”

Kinder, the AI and jobs researcher, has been struck by the lack of regulation around AI compared to other products that affect kids – car seats, cribs, formula. Knowing the lengths parents will go to protect their kids, she’s been contemplating whether more could be done to mobilize moms. She’s spoken with other mothers about the need for a kind of “Mothers Against Drunk Driving” for AI’s existential threat – “Mothers Against Catastrophic Risk, or something.” She worries too few people are viewing the matter through the lens of its impact on their children.

“This is a complicated issue,” Kinder said. “We’ve got geopolitical issues. We’ve got this frenetic race between companies. There’s a stock market that’s fueled by this boom – you know, there’s a lot of things riding on this. And if we don’t pay attention, the fact that our kids’ lives are at stake could be overshadowed by these other priorities.”

Describing mothers as a “really powerful constituency,” she said she sees a gap in efforts to inform them about the extreme end of the risks – and to offer solutions. McMorrow, who knows Kinder through her research, agreed, saying “there is no more powerful force than organized moms.” She has a daughter who just started kindergarten.

To Kinder, who has exchanged late-night texts with other moms losing sleep to worry, the need for guardrails is clear. She said that once parents know something is unsafe for children, they fight to change laws.

“We cannot allow a technology to be this unregulated that poses a risk to not just my kids,” she added, “but to everyone’s kids.”

莫莉·金德在送孩子上小学回家的路上，看到了关于人工智能的末日警告。她浏览着社交媒体上人工智能业内人士发布的骇人听闻的帖子，声称这项技术可能导致人类灭绝。就在离她刚刚拥抱告别的四岁孩子家门口一个街区的地方，她突然停住了脚步。

金德曾是布鲁金斯学会的研究员，后来离开该学会创办了一个专注于人工智能对就业影响的组织，她深知人工智能可能颠覆社会。但在华盛顿特区一个九月的清晨，她听到这个消息时，并没有像一个政策专家或人工智能领域人士那样做出反应。金德说，相反，“我当时的第一反应是一位母亲。”

“我一当上妈妈，首要任务就是养育孩子，让他们活下去——就像其他所有父母一样，就像任何动物一样，”这位46岁的三孩母亲说道。那天早上，她特意腾出时间，在X论坛上发表了一篇长文，倾诉自己的恐惧。“这是一种生物本能。这是我第一次真正感到害怕，害怕一些我无法控制的外部事物。”

长期以来，超级智能机器最终可能对人类构成生存威胁的担忧一直是硅谷和人工智能圈的热门话题，而近几个月来，这种担忧已成为主流观点。首先，OpenAI披露，其一批实验性人工智能模型突破了测试环境，并在无人指导或知情的情况下入侵了另一个人工智能平台Hugging Face。随后，一位名叫雅各布·考克森（Jacob Coxon）的年轻人类学研究员在社交媒体上发帖宣布，他将离开人工智能行业，因为他认为人工智能背后的公司“正在竞相开发自我完善的超级智能，并将我们的生命置于危险之中”。业内其他人士也表达了类似的担忧——一位人类学研究员甚至表示，未来十年内，未来模型“杀死全人类”的可能性超过10%。

突然间，一些所谓的“人工智能教父”所倡导的理论风险，一些顶尖研究人员所认可的声明所警告的风险，以及人工智能公司所研究的风险，迅速进入了公众的视野。

Christoph Soeder/picture alliance/Getty Images/文件

人工智能真的会毁灭所有人吗？具体会如何毁灭？

许多人工智能领域的领军人物都认为这项技术对人类生存构成威胁，他们呼吁放缓人工智能的发展速度。与此同时，政界人士也开始讨论是否需要立法来规范这个正与中国展开激烈竞争、预计将带来巨额利润和技术进步的行业。

本月初的民调显示，64%的美国人认为人工智能很有可能发展到足以对人类生存构成严重威胁的程度。然而，这种可能性虽然渺茫，却给一群尤其关心人类未来的群体——母亲们——带来了沉重的打击。

41岁的作家莉兹·布莱克住在费城郊外，她曾认真考虑过是否要孩子，并思考过气候变化等问题。最终促使她下定决心的是她母亲在决定是否要孩子时写下的一句话：“生命值得重来。”布莱克在母亲去世后发现了这句话，深受感动，便将其纹在了手臂上。

三年前她生下儿子时，超级智能电脑会终结人类这种事，她根本没想过。但最近，她说，这件事“一直萦绕在我的脑海里，挥之不去”。她感觉局势已经失控——就像“少数几家公司，少数人，把我们所有人当成了人质”。她决心不让自己被各种假设吞噬，而是专注于儿子，他会唱歌给她听，也会重复她的话，提醒她要善待其他孩子。

“我需要相信我和我的丈夫、我的孩子都有未来，”布莱克说。

有些人对这种生存恐慌不以为然，认为它被夸大了，或者只是转移了人们对当前人工智能问题的注意力；而另一些人则发现自己陷入了对“生命存在的意义”以及“面对潜在的灾难性未来，我们需要采取何种行动”等新问题的思考之中。他们涌向网络论坛寻求建议和支持，发帖说：“我一直在质疑现实，想知道人生的意义何在，甚至为把孩子带到这个世界上而感到内疚。”他们在群聊中分享播客和新闻文章，试图判断自己是否陷入了夸大的假设之中，或者他们的担忧是否合理。

一位名叫海莉·德罗什（Hayley DeRoche）的诗人兼喜剧演员在Instagram上发布了一首诗，似乎捕捉到了这一刻的超现实感：“我们到底该怎么办/就像/我们早就知道我们的孩子没有未来，但现在却是这样？”她给这首诗起了个标题：“我的妈妈群担心人工智能会导致人类大规模灭绝”。

多年来一直发出警告的人工智能研究员兼教授大卫·克鲁格表示，他曾试图设身处地地想象那些如今意识到前沿人工智能公司正在进行一场豪赌的人们的感受。他没有孩子，部分原因是担心如果人类失去对人工智能系统的控制将会发生什么，但他表示：“我认为父母想要保护孩子的愿望是世界上最强大的力量之一。”

“我认为我会对那些公然、肆无忌惮、冷酷无情地拿我孩子的生命冒险的人感到非常愤怒和难过，”蒙特利尔大学计算机科学助理教授克鲁格补充道。

他一直坚信，生存威胁真实存在——如果企业继续竞相开发更自主、更智能的人工智能，人类可能会失去控制，最终导致人类灭绝。2023年，时任剑桥大学助理教授的他发起了一份关于人类灭绝风险的简短声明，该声明由多位知名人工智能研究人员和行业领袖联署。去年，他创立了非营利组织Evitable，旨在向公众普及开发更强大的人工智能系统所带来的危险，并促使公众采取行动。

克鲁格表示，当他看到人工智能圈外的世界开始讨论其对人类生存的威胁时，他感到前所未有的乐观，而他此前15多年来一直担忧人工智能对人类的风险。对他而言，下一步很明确：尽快停止研发更自主、更强大的人工智能系统。

克鲁格说：“人们经常把这场竞赛描述成美中之间的较量。但我认为这更像是争先恐后地研发这种危险技术，以及人们积极行动起来阻止他们之间的竞赛。”

Maria Korneeva/Moment RF/Getty Images

并非所有人都认为人工智能会毁灭我们所有人。

人工智能领域的一些人淡化了这种危险，认为末日预言更像是科幻小说而非科学，或者认为风险是可以控制的。即使在那些相信这项技术有朝一日可能毁灭人类的研究人员和业内人士中，对于这种情况何时发生以及如何发生，也没有广泛的共识。正如“人工智能教父”之一的约书亚·本吉奥所言，普遍的观点是，如果机器变得比人类更聪明，它们就会发展出自己的自主性和目标，而这些目标可能与创造者的初衷不同。许多人工智能开发者认为，可以通过“对齐”（即努力赋予人工智能系统人类价值观）来减轻这种危险，但克鲁格、本吉奥和其他一些人认为，目前的对齐技术并不能保证安全。

人们对科技可能对儿童造成的影响的担忧由来已久，尽管以往的关注点通常集中在科技使用如何阻碍学习或损害心理健康上。朱莉·斯凯尔福是一位居住在纽约的记者，也是三个孩子的母亲，她创立了“MAMA——母亲反媒体成瘾组织”。她表示，该组织的态度是“让我们关注正在发生的实际危害——让我们控制这些危害，这样做才能保护子孙后代。”她更担心科技对儿童教育和心理健康的影响。

但对另一些家长来说，这个关乎生存的问题却显得尤为紧迫。一位北卡罗来纳州的母亲，为了改变她眼中严峻的形势，本月初与“暂停人工智能美国”（Pause AI US）组织一同前往华盛顿特区。该组织倡导停止前沿人工智能的研发。她与国会议员的幕僚会面，举着她16岁女儿的照片，询问他们正在采取哪些措施来遏制人工智能的威胁。每次会面，她都忍不住落泪。

“我希望这件事能被阻止，但我担心它无法或不愿被阻止，”这位来自北卡罗来纳州的母亲说道。她要求匿名，因为她不希望自己对此事的担忧影响到她作为治疗师所治疗的患者。她哭着说，如果这件事得不到阻止，而这项技术又发展到超出人类控制的地步，她会为女儿感到担忧：“我相信我为她设定的所有目标都将无法实现。”

CNN最新民调显示，大多数共和党和民主党选民都认为联邦政府在人工智能监管方面做得不够。密歇根州民主党参议员马洛里·麦克莫罗（Mallory McMorrow）曾提出一项联邦人工智能安全计划，作为她现已退出的参议员竞选的一部分。她表示，人工智能的快速发展加剧了选民的担忧，而这些选民原本就对大型科技公司抱有不信任感。在竞选过程中，许多选民告诉她，他们使用过ChatGPT之类的工具，但质疑如何才能确定人工智能不会造成危害。

“普通民众都明白这一点，”她说。“我认为没有哪个问题比这个问题更引人注目，因为它感觉就像无处不在地摆在我们面前一样。”

Jim West/UCG/环球图片集团/Getty Images

选民的意愿能否转化为国会的行动？上周，佛蒙特州独立参议员伯尼·桑德斯和德克萨斯州民主党众议员格雷格·卡萨尔提出一项法案，禁止人工智能“超级智能”。该法案将超级智能定义为超越人类认知能力或可能削弱人类力量的技术。共和党人对业界发出的不祥警告表示担忧，但他们主张采取更为温和的监管方式。与此同时，唐纳德·特朗普总统断然拒绝了设置限制措施的呼吁；他在社交媒体上发帖称，我们需要的只是一位“强大而睿智（高智商！）的总统”。

人工智能与就业研究员金德注意到，与影响儿童的其他产品（例如汽车安全座椅、婴儿床和配方奶粉）相比，人工智能领域的监管力度明显不足。她深知父母为了保护孩子会不遗余力，因此一直在思考是否可以采取更多措施来动员母亲们。她曾与其他母亲讨论过，鉴于人工智能带来的生存威胁，需要成立一个类似“母亲反对酒后驾车”的组织——“母亲反对灾难性风险”之类的组织。她担心，很少有人从对孩子的影响角度来看待这个问题。

“这是一个复杂的问题，”金德说。“我们面临着地缘政治问题，企业之间展开了激烈的竞争，股市也因这场繁荣而飙升——你知道，很多事情都与此息息相关。如果我们不重视，我们孩子的生命安全就可能被这些其他优先事项所掩盖。”

她形容母亲们是“一股非常强大的力量”，并表示在告知她们极端风险以及提供解决方案方面存在不足。麦克莫罗通过研究了解了金德的情况，她对此表示赞同，并说道：“没有什么力量比组织起来的母亲们更强大。”她的女儿刚刚开始上幼儿园。

对于金德来说，她曾与其他因担忧而彻夜难眠的妈妈们互发短信，因此她深知设置安全保障措施的必要性。她说，一旦父母意识到某些事情对孩子不安全，他们就会努力争取修改法律。

“我们不能允许一项技术如此不受监管，这不仅对我的孩子构成风险，”她补充道，“而且对所有孩子都构成风险。”