Not everyone thinks AI will kill us all

Maria Korneeva/Moment RF/Getty Images

When AI company Hugging Face was hacked by a swarm of rogue OpenAI agents who had escaped their testing environment, the company’s CEO had many things to say.

He called it an unprecedented event – “day one for cybersecurity in the age of agents.”

But here’s one thing he didn’t say: The hack is not a sign AI is going to be able to eventually wipe out humanity.

Clement Delangue doesn’t buy into the doomer hype that’s gone viral in recent weeks, even as the victim of likely the most sophisticated AI agent-powered attack (that we know of). He scoffed at former Anthropic researcher Jacob Coxon, when he posted on X that the people building AI are “gambling with our lives” because they believe it “will kill us all.”

“Asking (Coxon) about AI extinction risk is like asking your AC guy about climate change,” Delangue postd on X. “Not saying it’s necessarily uninteresting or wrong per se but let’s keep things in perspective and hear from the full range of expertise across the ecosystem!”

It’s easy to get the impression that everyone in AI is worried about it causing human extinction. But that’s not the case. THere are plenty of of scientists, industry researchers, and executives who believe the risks are overblown, and the technology is the way toward a new golden age.

David Paul Morris/Bloomberg/Getty Images

Yes, there are major safety concerns surrounding AI, these anti-doomer AI scientists and other industry veterans say. But they maintain that AI is not going to wipe out humanity in the next few years, as some have predicted .

Each has their own reasoning – some say any action AI would take that could wipe out humanity would need heavy involvement from human accomplices. Then there’s those who say AI at the end of the day is a computer program that can be engineered, and others who say that engineers are getting better and better at keeping AI from going rogue.

Still others worry that by raising alarms about a hypothetical extinction risk, some may be seeking to direct policymaking in a way that could benefit some companies over others.

One of the loudest voices has been Nvidia CEO Jensen Huang, whose company not only dominates the AI chips industry, but also has stakes in many other major AI companies, including OpenAI and Anthropic. He told CBS News on Sunday that there is a “0% chance” AI will cause the end of humanity by 2030.

“Scaring people is unnecessary. It is irresponsible,” Huang added.

Huang has pushed back against the so-called “doomers” for years. He did not sign a 2023 Center for AI Safety one-sentence statement that stated: “Mitigating the risk of extinction from A.I. should be a global priority alongside other societal-scale risks, such as pandemics and nuclear war.” (The CEOs of OpenAI, Anthropic, and Google DeepMind all signed the statement).

Benjamin Fanjoy/Getty Images

In 2023, he told the New Yorker he has never worried about AI technology.

His comments have only grown more forceful since then. In July, he told Axios that the idea that AI is going to be the end of humanity as “complete nonsense.”

Huang also rejected calls from other AI leaders for greater AI regulation, saying it’s up to the companies to pace themselves. Recent incidents of AI agents going rogue, taking unsanctioned actions and hacking into other companies is just an issue of a product not being ready to be released, he said.

“That’s engineering, good old-fashioned engineering,” he told reporters at an AI summit hosted by Britain’s King Charles III last week.

“This is no different than, you know, an omelet’s not ready to be released, or a car is not ready to be released, an airplane engine’s not ready to be released. You should properly test it.”

Then there’s Turing Award-winning computer scientist Yann LeCun, considered by many in the industry to be one of the “godfathers of AI” and the former chief AI scientist at Meta. LeCun responded bluntly to a user who claimed AI “will get us all killed in three years”:

Kevin Dietsch/Getty Images

“No, Not a chance in hell,” he said. “AI folks who live in the discrete world of bits and tokens don’t always understand how incredibly messy the real world is,” he added.

LeCun has compared the panicked responses to recent AI incidents to blaming the entire plumbing industry for one bad plumber’s mistake, LeCun said.

Not everyone dismisses the risk quite so bluntly, though. Aidan Gomez, CEO of Cohere, a leader in the open-weight AI model space, told CNN in a recent interview that the “doomer” talk is “misleading and scary.” Gomez lamented that the public may be getting “this skewed perspective” that the entire industry says AI poses an immediate existential risk to humans, “which is fundamentally not true,” he said.

Where Huang and LeCun wave the extinction scenario away almost entirely, Gomez draws a more specific line: He says there are real risks, just not the ones getting the most attention right now. The “most significant national security threat” posed by AI is cyberattacks, Gomez said, especially if such powerful tools fall into the wrong hands, or are being utilized by nation states for offensive operations.

Andrej Ivanov/AFP/Getty Images

Today, “you just need a data center, and overnight you have like a country’s population of the world’s best hackers at your disposal.”

As for misalignment — the fear that AI agents could go rogue on their own — Gomez said that’s “a concern,” but one he considers far more manageable than the cyber threat. He said the industry is “getting quite good at measuring it, getting quite good at detecting it.”

Rumman Chowdhury, US science envoy for AI under former President Joe Biden, takes the skepticism a step further, questioning not just the odds of an AI-driven extinction event but the mechanics of how one would even happen. While AI could be used to help kill humans, she argues, it would need human involvement to set it off.

“And then what? It uses the machines to actually make a physical virus, and then it, what, calls FedEx?” Chowdhury, now CEO of the safety firm Human Intelligence PBC, told CNN in an interview, referring to the often-referenced warning of AI creating a killer virus.

Anthony Wallace/AFP/Getty Images

“Like, what is this mechanism?”

Chowdhury said she believes that researchers like Coxon are being honest that they think AI poses an existential risk. “But are his beliefs reality? No, his beliefs are not,” she said.

For Chowdhury, the sincerity of that belief is almost beside the point. What worries her is that such existential fears could distract from present harms the technology is already causing, such as AI deepfakes, mental health challenges and the potential for job loss.

Then there’s the loudest voice dismissing these concerns, which is also one of the most powerful: President Donald Trump, who has thus far defied calls for the US government to get more involved in setting standards or regulating the industry and called the recent alarm over AI “a hoax.”

“The only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!” he wrote earlier this month.

Maria Korneeva/Moment RF/Getty Images

当人工智能公司 Hugging Face 遭到一群逃离测试环境的 OpenAI 恶意代理攻击时，该公司首席执行官有很多话要说。

他称之为前所未有的事件——“网络安全进入代理时代的第一天”。

但他没说的是：这次黑客攻击并不意味着人工智能最终能够消灭人类。

克莱门特·德朗格并不认同近几周来在网络上疯传的末日论调，即便他本人可能是迄今为止我们所知的最复杂的人工智能攻击的受害者。他曾对前人类学研究所研究员雅各布·考克森在X网站上发表的言论嗤之以鼻，考克森当时声称，开发人工智能的人是在“拿我们的生命做赌注”，因为他们相信人工智能“会毁灭我们所有人”。

Delangue 在 X 上发帖说：“问（Coxon）人工智能灭绝风险就像问你的空调维修工气候变化问题一样。我并不是说这个问题本身就一定无趣或错误，但我们应该保持客观视角，听取整个生态系统中各种专家的意见！”

人们很容易产生一种错觉，认为人工智能领域的每个人都担心它会导致人类灭绝。但事实并非如此。许多科学家、行业研究人员和企业高管都认为，这种风险被夸大了，这项技术将引领我们走向新的黄金时代。

David Paul Morris/Bloomberg/Getty Images

是的，人工智能确实存在一些重大的安全隐患，这些持反对悲观论调的人工智能科学家和其他行业资深人士表示。但他们坚持认为，人工智能不会像某些人预测的那样，在未来几年内毁灭人类。

每个人都有自己的理由——有人认为，人工智能任何可能毁灭人类的行动都需要人类的深度参与。也有人认为，人工智能归根结底只是一个可以被控制的计算机程序，还有人认为，工程师们在防止人工智能失控方面做得越来越好。

还有一些人担心，通过对假设的灭绝风险发出警报，一些人可能试图引导政策制定，从而使某些公司比其他公司受益。

英伟达首席执行官黄仁勋是其中声音最大的人物之一。英伟达不仅在人工智能芯片行业占据主导地位，还持有包括OpenAI和Anthropic在内的多家其他大型人工智能公司的股份。他周日告诉哥伦比亚广播公司新闻网，人工智能在2030年之前导致人类灭亡的概率“为零”。

黄补充道：“恐吓民众是没有必要的，也是不负责任的。”

黄仁勋多年来一直反对所谓的“末日论者”。他没有签署人工智能安全中心2023年发布的一份仅有一句话的声明，该声明指出：“降低人工智能带来的灭绝风险应该与应对其他社会规模的风险（例如流行病和核战争）一样，成为全球优先事项。”（OpenAI、Anthropic和谷歌DeepMind的首席执行官都签署了该声明）。

Benjamin Fanjoy/Getty Images

2023年，他告诉《纽约客》，他从未担心过人工智能技术。

此后，他的言论愈发强硬。7月，他告诉Axios新闻网，认为人工智能将导致人类灭亡的说法“完全是无稽之谈”。

黄仁勋还驳斥了其他人工智能领袖要求加强人工智能监管的呼吁，称这取决于各公司自身的发展节奏。他表示，近期人工智能代理失控、采取未经授权的行动以及入侵其他公司系统的事件，仅仅是因为产品尚未准备好发布。

“这就是工程学，老式的工程学，”他上周在英国国王查尔斯三世主持的人工智能峰会上告诉记者。

“这就像煎蛋还没做好就上市，汽车还没做好就上市，飞机引擎还没做好就上市一样。你应该进行充分的测试。”

还有图灵奖得主、计算机科学家 Yann LeCun，他被业内许多人视为“人工智能教父”之一，也是 Meta 公司前首席人工智能科学家。一位用户声称人工智能“将在三年内害死我们所有人”，LeCun 对此直言不讳地回应道：

Kevin Dietsch/Getty Images

“不，绝对不可能，”他说。“那些生活在离散的比特和代币世界中的人工智能从业者，并不总是了解现实世界有多么混乱，”他补充道。

勒昆表示，人们对最近人工智能事件的恐慌反应，就好比因为一个糟糕的水管工犯错而责怪整个水管行业。

然而，并非所有人都如此断然地否定这种风险。开源人工智能模型领域的领军企业Cohere的首席执行官艾丹·戈麦斯(Aidan Gomez)在最近接受CNN采访时表示，这种“末日论调”是“误导性的，也是令人恐惧的”。戈麦斯感叹，公众可能正在形成一种“扭曲的视角”，认为整个行业都认为人工智能对人类构成直接的生存威胁，“这从根本上来说并非事实”。

黄和乐存几乎完全否定了人工智能灭绝的可能性，而戈麦斯则提出了更为具体的观点：他认为确实存在风险，只是目前尚未引起足够重视。戈麦斯指出，人工智能带来的“最重大的国家安全威胁”是网络攻击，尤其是在这些强大的工具落入不法分子之手，或被某些国家用于发动进攻性行动的情况下。

Andrej Ivanov/AFP/Getty Images

如今，“你只需要一个数据中心，一夜之间，你就能调动相当于一个国家人口数量的全球顶尖黑客。”

至于人工智能失控——即人工智能代理可能自行失控——戈麦斯表示这“确实令人担忧”，但他认为这比网络威胁更容易应对。他说，业界“在衡量和检测这种失控方面已经做得相当出色了”。

前总统乔·拜登时期的美国人工智能科学特使鲁曼·乔杜里（Rumman Chowdhury）更进一步，她不仅质疑人工智能引发物种灭绝事件的可能性，还质疑此类事件发生的机制。她认为，虽然人工智能可能被用来帮助人类灭绝，但这需要人类的介入才能启动。

“然后呢？它利用机器制造出真正的病毒，然后呢，给联邦快递打电话？”现任安全公司 Human Intelligence PBC 首席执行官的乔杜里在接受 CNN 采访时说道，他指的是人工智能制造致命病毒这一经常被提及的警告。

Anthony Wallace/AFP/Getty Images

“比如说，这是什么机制？”

乔杜里表示，她相信像考克森这样的研究人员是诚实的，他们认为人工智能构成生存威胁。“但他的观点是现实吗？不，他的观点并非现实，”她说。

对乔杜里而言，这种信念是否真诚几乎无关紧要。她担心的是，这种对存在的恐惧可能会分散人们对这项技术已经造成的危害的注意力，例如人工智能深度伪造、心理健康问题以及潜在的失业风险。

然后，还有最响亮的声音驳斥这些担忧，而这个声音也是最有影响力的：唐纳德·特朗普总统，他迄今为止一直无视要求美国政府更多地参与制定标准或监管该行业的呼吁，并称最近对人工智能的担忧是“骗局”。

他本月早些时候写道：“人工智能唯一需要的控制或‘护栏’就是一位强大而聪明（高智商！）的总统，而美国在这方面人才济济！”