OpenAI reportedly kicked around an 'insane' plan to pit world leaders against each other like a Call of Duty villain

The New Yorker has published an enormous feature about OpenAI CEO Sam Altman and his disputed trustworthiness, citing more than one person who has accused the generative AI mogul of habitual dishonesty.

The 16,000 word article (available online or in The New Yorker's latest print edition) adds new context to a number of widely-reported episodes from Altman's career, including his 2023 ousting from OpenAI and return, his beef with Elon Musk, and the disintegration of his persona as a humanity-first AI safety advocate, which strained credulity from the start and is now especially comical in juxtaposition with his current role as a profit-seeking Trump ally who recently signed a deal with the US Department of War.

One unflattering portion of the article, among many, discusses a defunct plan to pit world leaders against each other by positioning OpenAI as a kind of nuclear weapon that they'd better compete to invest in, lest they be left behind. OpenAI denies that characterization of the discussions, calling it "ridiculous," but former OpenAI employees say otherwise.

After former OpenAI policy adviser Page Hedley presented ways to avoid a global AI arms race, OpenAI president and major Trump donor Greg Brockman reportedly proposed the opposite. In The New Yorker's words, the idea was that "OpenAI could enrich itself by playing world powers—including China and Russia—against one another, perhaps by starting a bidding war among them."

Jack Clark, who had been OpenAI's policy director when the plan was discussed and is now head of policy at competitor Anthropic, described it as "a prisoner’s dilemma, where all of the nations need to give us funding" which "implicitly makes not giving us funding kind of dangerous."

OpenAI says that no such plan was taken seriously, and that at most "ideas were batted around at a high level about what potential frameworks might look like to encourage cooperation between nations."

But The New Yorker, which says it reviewed documents from the time, says that the "countries plan" was real, was popular with OpenAI executives, and was only abandoned after employees discussed quitting over it.

A junior researcher told the publication that, during a meeting in which the plan was discussed, they thought: "This is completely fucking insane."

The big article—here's the link again—contains many other details about OpenAI's part in the great AI bubble and Altman's reputation among his peers, and is a decent way to pass some time while you wait for the bubble to pop so you can buy RAM again. (It may be wishful thinking on my part that RAM prices will ever go back to normal, but one has to have hope.)