Palisade Research, a nonprofit that studies AI capabilities and motivations, has released a dozen video interviews with current and former employees of OpenAI, Google DeepMind, and Anthropic on a new site, frominside.ai. The project gives a platform to industry insiders who are publicly warning about the technology they help build.
Several researchers describe existential risk in stark terms. Geoffrey Irving, formerly of OpenAI and Google DeepMind, says the chance of human extinction is 'about a coin flip.' Google DeepMind's Neel Nanda puts the chance of AI-caused extinction at at least 10 percent, which he calls 'ridiculously high.' Former OpenAI researcher Daniel Kokotajlo says superintelligent AI would be 'god-like powerful' and that nobody would be able to control it. The speakers differ on the odds, and the videos offer no single solution.
The researchers also address their own credibility. Google's Mary Phuong says viewers should be suspicious because she is paid by a lab, while Nanda says he would quit if he did not think his work reduced existential risk, but that companies would continue building systems regardless. The Verge notes that the interviews are part of a broader and unsettled conversation about what AI safety means and what should be done about it.