Hinton warns that a Chernobyl-level accident could occur in the AI field this year or next
English text is being prepared.
In an interview, Hinton said that a Chernobyl-level accident could occur in the AI field this year or next, and compared this summer's incident in which an OpenAI agent breached the Hugging Face platform to the Three Mile Island accident in the history of nuclear safety, considering it a dangerous warning but not yet a true disaster. The Chernobyl of the AI world he refers to is not AI rebelling on its own, but AI being exploited by hackers to attack critical infrastructure, for example causing widespread, prolonged power outages or paralyzing the global financial system for a week. He named and criticized the risk attitudes of the leadership at leading AI companies such as OpenAI's Altman, Meta's Zuckerberg, and Musk, while stating that he is not opposed to AI and recognizes its positive value.
Why read this
AI score 60
Hinton uses Three Mile Island as an analogy for current AI risks and lays out the specific disaster scenarios he fears, so readers can understand the focus of the risk discussion.
Topics
Body
辛顿把今年夏天OpenAI智能体入侵Hugging Face平台事件，类比为核安全史上的三里岛事故，属于一次危险的预警信号，但这还不是真正的灾难。
所谓AI界的“切尔诺贝利”，并不是科幻电影里AI自主反叛、消灭人类，而是AI被黑客利用，发起针对关键基础设施的网络攻击，最典型的场景就是造成大范围长时间停电，或是让全球金融系统瘫痪一周，直接冲击民生运转秩序。
在采访中，辛顿直言自己并不看好头部AI企业管理层的风险态度，点名批评OpenAI的奥特曼、Meta的扎克伯格、马斯克等人。
需要指出的是，辛顿并非反对AI，他认为AI就是一把双刃剑，自己认可其拥有巨大正向价值，能够给人类带来美好的未来，可一旦重大事故发生，人们或许会后悔当初不该发明这项技术。