The debates

A coalition says Anthropic is not a safe AI company

The Center for AI Safety and signatories ranging from Steven Pinker and Judea Pearl to Guillaume Verdon and three Fields medallists said on 18 November that Anthropic races to build the recursively self-improving AI it warns about.

Filed under AI safety, AI industry

The statement, timed for the IPO, accuses Anthropic of using safety branding to court investors and staff while weakening its own safety pledges and lobbying for regulation that suits it. Dan Hendrycks, the centre’s executive director, said the signatories disagree on almost everything else. Peter Scholze, James Maynard and David Mumford signed, as did Max Tegmark, Lawrence Lessig and Thomas Wolf of Hugging Face.

Two days later Alpöge addressed the timing of his announcement, a day after the listing, in a note to his proof. He wrote that the result came late on a Sunday, that there was a constraint to respect before he could announce, and that the cynical reading of the timing is false.