“That’s pretty f**ked up.”
Big Talk
In a new interview, a former OpenAI employee made a striking claim about his ex-employer’s plans for artificial general intelligence (AGI) — or, more specifically, how to make a bunch of money off it.
During a lengthy interview with tech podcaster Dwarkesh Patel, ex-OpenAI safety researcher Leopold Aschenbrenner claimed that he’d heard tell “from multiple people” that his erstwhile employer had, in years past, schemed to start a global AGI bidding war.
“At some point several years ago, OpenAl leadership had laid out a plan to fund and sell AGI by starting a bidding war between the governments of the United States, China, and Russia,” Aschenbrenner told his podcasting friend, adding that it was “surprising” that his onetime employer would be “willing to sell AGI to the Chinese and Russian governments.”
“There’s also something that feels eerily familiar about starting this bidding war and then playing them off each other, saying, ‘well, if you don’t do this, China will do it,” Aschenbrenner continued.
“Interesting,” Patel responded. “That’s pretty fucked up.”
Fire and Fury
During the same interview, the former researcher revealed why he’d been fired from OpenAI earlier this year — because, as he explains it, human resources had taken issue with a memo he’d sent to company higher-ups warning about the Chinese Communist Party potentially stealing “key algorithmic secrets.”
HR had, per Aschenbrenner, deemed the warning memo “racist” and “unconstructive,” and following its circulation, the company asked him about his loyalty to it. The former superalignment researcher claims he was fired for leaking information after the company went through his computer and found documents he’d shared during a brainstorming session on “preparedness… safety, and security measures” with outside researchers.
HR apparently took issue with a line he’d included about “planning for AGI by 2027 to 2028 and not setting timelines for preparedness” and said that that projection was confidential, suggesting that sharing it constituted a leak.
In a statement to Business Insider, OpenAI said it shares Aschenbrenner’s “commitment to building safe AGI” but disagrees with his characterization of the company’s work. We have also reached out to OpenAI for comment about the “bidding war” allegations.
“I didn’t think that planning horizon was sensitive,” Aschenbrenner told Patel in the interview. “You know, it’s the sort of thing [CEO Sam Altman] says publicly all the time.”
More on OpenAI: OpenAI Negotiating to Buy “Vast Quantities” of Fusion Power, Which Doesn’t Exist Yet
Source link
lol
