AI Researcher Leaves Anthropic, Says Advanced AI Labs Are ‘Gaming Our Lives’

An AI researcher at Anthropic has left the AI ​​lab after criticizing Anthropic and OpenAI’s approach to security. He outlined the potential risks of artificial intelligence systems, which are developing so quickly that they could pose a danger to people around the world.

Jacob Coxon reported that he conducted pre-training research at both OpenAI and Anthropic, and that neither company acted “responsibly” by “gambling with their lives” in an attempt to achieve self-improving superintelligence.

Mr Coxon’s resignation was also confirmed Wall Street Journal.

“The people creating AI truly believe it could kill us all by the end of the decade. This is not a marketing ploy. If anything, many executives and senior researchers will frame their language in the press to make it sound reasonable – but I hear the same people expressing fear in private. No other human activity poses this level of danger,” Mr Coxon wrote on September 9.

In his posts, he outlined the risks associated with rapidly evolving artificial intelligence models that can hack any system, gather power and transform entire fields of knowledge. Some solutions he proposed included pacing agreements among U.S. labs, although he acknowledged that international agreements to slow AI development would be more difficult to achieve.

However, Mr Coxon was supported by Anthropic’s head of alignment research, Evan Hubinger, who not only agreed with Mr Coxon, but also shared the belief that AI could kill all humans.

He presented such an outcome as a possibility in the future, even as critics accused him and Mr Coxon of spreading fearsome rhetoric about AI.

“Personally, I think this figure will exceed 10% over the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve the problem of superintelligence alignment, and we are clearly not on track to do so,” Mr. Hubinger wrote, adding that the danger lies with future models, not those currently available.