Hugging Face CEO says asking Anthropic researcher ‘about AI extinction risk is like asking your AC guy about climate change’

Hugging Face CEO Clement Delangue questioned the attraction connected a erstwhile Anthropic interrogator who said AI workers privately fearfulness the exertion could termination each humans.

"Sorry, but asking Jacob astir AI extinction consequence is for illustration asking your AC feline astir ambiance change," Delangue wrote connected X connected Friday.

Delangue said that Jacob Coxon, the erstwhile Anthropic interrogator whose resignation has sparked a renewed statement astir AI's risks, should beryllium put into the due context.

"Not saying it's needfully uninteresting aliases incorrect per se but let's support things successful position and perceive from the afloat scope of expertise crossed the ecosystem!" Delangue wrote.

Coxon wrote connected Tuesday connected X that he was resigning owed to concerns that some Anthropic and OpenAI, wherever he antecedently worked, are "racing consecutive to self-improving superintelligence and gambling pinch our lives."

"The group building AI earnestly judge that it could termination america each by the extremity of the decade," he wrote. "This is not a trading stunt."

Delangue later said that his comparison to an "AC guy" was not meant "to discredit" Coxon "at all." Rather, Delangue said, location needs to beryllium a much robust speech astir AI's risks.

"It's simply that we should put immoderate 1 position successful discourse and perceive to the broader AI ecosystem too," he wrote.

Delangue besides said it was his "understanding" that Coxon mostly focused connected pretraining, successful which researchers train AI models connected ample datasets, alternatively than much risk-focused research.

In his bid of posts, Coxon said he worked connected pretraining investigation astatine some OpenAI and Anthropic "for the past 3 years." Ethan Perez, alignment squad lead astatine Anthropic, later wrote connected X that Coxon was a "senior interrogator who joined Anthropic successful May" and that OpenAI's main rival had been trying to enlistee Coxon for astir 2 years.

Some successful tech, for illustration SpaceX CEO Elon Musk, whose institution includes SpaceXAI, person been captious of Coxon's claims. In consequence to personification who recounted the lead-up to The Wall Street Journal's first communicative connected Coxon's resignation (which came earlier Coxon's tweets), the SpaceX CEO wrote that "the groundwork for this psy op (for deficiency of a amended term) has been prepared for a agelong time."

Others, including researchers astatine Anthropic and OpenAI, person said Coxon's claims astir the consequence of AI wiping retired humanity are well-founded.

"Jacob is correct here—we really do earnestly judge AI could termination each humans!" Evan Hubinger, the alignment subject lead astatine Anthropic, wrote connected X. "I personally deliberation it is >10% wrong the adjacent decade."

As for different perspectives that mightiness beryllium helpful, Delangue wrote successful a reply that includes, "social scientists, different AI builders, futurists, investors, policymakers."

Delangue and Hugging Face's profiles person exploded since OpenAI announced successful July that rogue AI models had surgery containment during cybersecurity testing and breached Hugging Face successful an effort to cheat connected soul tests.

On September 3, Nvidia, which is highly reliant connected AI-related request for its chips, announced plans to acquire Hugging Face for nearly $13 billion. If policymakers were to slam the brakes connected frontier AI improvement (a telephone immoderate person renewed successful the aftermath of Coxon's comments), Nvidia would apt beryllium importantly affected.