DeepMind chief Demis Hassabis calls for US-led body to test ‘frontier’ AI models - FT中文网
登录×
电子邮件/用户名
密码
记住我
请输入邮箱和密码进行绑定操作:
请输入手机号码,通过短信验证(目前仅支持中国大陆地区的手机号):
请您阅读我们的用户注册协议隐私权保护政策,点击下方按钮即视为您接受。
FT商学院

DeepMind chief Demis Hassabis calls for US-led body to test ‘frontier’ AI models

Society has ‘precious window’ to prepare for technology advancing at historic speed, warns Nobel laureate
00:00

{"text":[[{"start":9.2,"text":"Google DeepMind chief executive Demis Hassabis has called for the creation of a US-led standards body to test new “frontier-class” AI models for national security threats, arguing that “urgent action” from international regulators is needed to address the risks posed by the rapidly advancing technology."}],[{"start":27.7,"text":"The warning from Hassabis, a Nobel laureate who leads Google’s AI efforts,"}],[{"start":33.05,"text":"follows the White House’s abrupt export ban on Anthropic’s most advanced models last month, alongside a fresh wave of warnings about the potential for AI to disrupt the global economy and financial system. "}],[{"start":45.65,"text":"The intervention is the most detailed proposal yet for AI regulation from Google, which is vying for AI leadership with Anthropic and OpenAI. "}],[{"start":53.849999999999994,"text":"“We’ve already seen the challenges frontier models pose for cyber security, and other threats including nuclear and bio risks may soon emerge as capabilities continue to advance,” Hassabis wrote in an essay published on social media app X on Tuesday."}],[{"start":69.44999999999999,"text":"“The rapid progress we’re seeing in AI requires a new approach to testing frontier AI model capabilities that is dynamic, adaptable and rigorous. The US is well positioned, given its economic and technical standing, to take the first step in developing such a framework.”"}],[{"start":87.89999999999999,"text":"Hassabis proposes that the new body could be modelled on the Financial Industry Regulatory Authority (Finra), the US-based self-regulatory organisation that oversees capital markets to protect investors. "}],[{"start":100.69999999999999,"text":"AI models designated as “frontier-class” — including both open and closed systems developed anywhere — would be vetted. He suggested the body should also stress-test AI models’ safety guardrails. A warning from Amazon that guardrails on Anthropic’s powerful Mythos model could be bypassed helped to prompt the US ban."}],[{"start":121.74999999999999,"text":"“This US-initiated effort would provide a strong starting point for creating shared international standards on Frontier AI,” Hassabis said, adding that it should then “spur the international community to reach a consensus” on AI regulation."}],[{"start":136.79999999999998,"text":"Hassabis’s move follows escalating calls by Anthropic chief executive Dario Amodei for stronger AI regulations, including allowing the government to block deployment of technology that is deemed unsafe. Last month, OpenAI also agreed to stagger the release of its latest model, GPT-5.6, following requests from the US government. "}],[{"start":158.7,"text":"Writing in the FT this month, OpenAI chief Sam Altman proposed the creation of “a US-led international forum that establishes accepted standards, provides expert and impartial analysis of capabilities and risks”, citing the International Atomic Energy Agency as a model. "}],[{"start":175.6,"text":"Tech industry executives at the G7 meeting last month urged world leaders to take a united front on AI safety and security, after the US government’s action against Anthropic prompted fears in Europe and Asia that they could be cut off from a vital engine of economic growth and military advantage. "}],[{"start":193.04999999999998,"text":"Discussion of AI regulation in the US has also intensified as Chinese capabilities improve. The latest open models from Z.ai and DeepSeek are narrowing the gap with their US rivals."}],[{"start":205.35,"text":"Beyond the regulatory proposals, Hassabis’ essay painted an optimistic future for AI, which he said could yield huge advances in medicine and energy. Artificial general intelligence that can match all the capabilities of the human brain was “probably only a few short years away”, he added, offering a “precious window” for society to act. "}],[{"start":224.45,"text":"“The magnitude of this technology’s impact will be unprecedented, perhaps 10x of the Industrial Revolution at 10x the speed,” he wrote. “If you stop to think about it, we’ve essentially found a way to make sand think. It’s miraculous.”"}],[{"start":248.25,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1784295924_6573.mp3"}

版权声明:本文版权归FT中文网所有,未经允许任何单位或个人不得转载,复制或以任何其他方式使用本文全部或部分,侵权必究。

欧洲风机制造商探讨合并,抗衡中国对手

欧盟更新并购审查指南之际,中国竞争压力或将推动行业并购。

特朗普压低油价的能力受到考验

黑格:随着伊朗冲突一拖再拖,口头干预或已不足以抑制油价。

朝韩竞相建造核动力潜艇

去年获得唐纳德•特朗普首肯后,韩国看到了发展更强大潜艇的机会。

电子价签之争凸显美国经济焦虑

新泽西州成为首个叫停电子价签的州,原因是担心这些价签会将动态定价机制引入杂货店,甚至监视顾客。

一周展望:美联储会加息吗?

投资者本周还将密切关注英国央行可能如何应对通胀再度抬头的迹象。

使用Kimi K3等中国开源AI模型有何风险?

萨克斯:真正的问题不在于海外开源,而在于缺乏保护基础设施的协调机制来应对网络攻击。
设置字号×
最小
较小
默认
较大
最大
分享×