Sam Altman Reportedly Laid Out Three Endgames For AI, And One Ends With Rivals’ GPUs ‘Melted’

OpenAI CEO Sam Altman reportedly told staff last week that the industry has discussed three ways to get through the race to superintelligence safely, and one of them ends with a single company “melt[ing] everyone else’s GPUs,” The Information reported on Monday.

The account rests on anonymous sources, but it lands in the backdrop of a year when the AI conversation has reached a fever pitch, and AI products are already being actively informed by Washington as well as by the labs.

Three plans

According to The Information’s reporting, the first is an international body, modeled on the International Atomic Energy Agency, that would enforce safety standards on frontier labs. Two people with knowledge of the remarks told The Information that Altman appeared to support it. The second, which he called the “Demis plan” after Google DeepMind’s Demis Hassabis, would fold the major labs into a single CERN-style entity that supplies models to partner countries.

The third is the “Eliezer plan,” after the AI safety theorist Eliezer Yudkowsky. A developer reaches self-improving AI first and uses the resulting economic power to starve rivals of chips. “Obviously I think that one’s terrible,” Altman apparently said, adding that he included it for “completeness.” The Information says a decision is due within a year.

Not a new idea?

Plan one has a long paper trail. In May 2023, Altman, OpenAI president Greg Brockman and then-chief scientist Ilya Sutskever wrote that the industry would likely need something like an IAEA for super intelligence efforts, able to inspect systems, require audits and track compute and energy use.

“As a first step, companies could voluntarily agree to begin implementing elements of what such an agency might one day require, and as a second, individual countries could implement it,” they wrote.