Robert O’Callahan, a Google DeepMind engineer based in Auckland, has resigned over how fast artificial intelligence is advancing. In a resignation email shared on his blog this week, he wrote that AI progress is “far too rapid”. His team’s goal of making AI “much cheaper and lower-latency”, he added, was not good for people right now.Yet O’Callahan stops short of the extinction warnings that have dominated the AI debate this month. He wrote that he is “not convinced the chance of ASI doom is 100%”, referring to artificial superintelligence. The risk, in his view, is “real but uncertain”. That alone, he argued, makes chasing ASI in the near future “inherently irresponsible”.
Google DeepMind engineer says his chip design work was speeding up AI
O’Callahan did not build AI models himself. He worked on tools for hardware chip design. For a while, he wrote, he told himself the work was relatively harmless. Over time he came to accept that its main impact would be faster design of a new breed of AI chips, making AI cheaper, faster and more capable.His exit lands in the same week Google laid out plans to test its AI chips in space. Under Project Suncatcher, a prototype satellite carrying Google’s Trillium TPUs will fly on SpaceX’s Transporter-18 rideshare mission. The satellite was built with Earth imaging company Planet. The long-term goal is to see whether orbit could host AI data centres, since solar panels there can generate up to eight times more power than on Earth. “One small step for TPUs…,” CEO Sundar Pichai posted on X. O’Callahan’s post does not mention the project.Leaving was not easy. He said being “paid handsomely to solve fun puzzles” had been amazing. Moving teams was impractical too, since Google has no other engineering teams in New Zealand. A Christian and a church elder in Auckland, he wrote that ignoring his work’s impact “would not be a Jesus-following thing to do”.
This AI safety resignation stops short of predicting doom
He joins a string of researchers who have left top AI labs over safety fears. Jacob Coxon quit Anthropic earlier this month, later warning in an interview that “we could all die” if development did not slow down. Former Google DeepMind researchers Rishub Jain and Bilal Chughtai also left this year.Anthropic CEO Dario Amodei has since proposed slowing frontier AI development, with Sam Altman, Demis Hassabis and Elon Musk backing him publicly. Nvidia CEO Jensen Huang pushed back, calling extinction warnings “doomsday narratives”.O’Callahan sits somewhere between the two camps. He wrote that several doomer predictions, including reward hacking, misalignment and deceptive models, have already come true. But he has seen no hard evidence that insiders are hyping AI risk to lift stock prices or shut out rivals. Regulation, in his words, is “desperately needed”.
O’Callahan plans to keep working on AI after leaving Google
He admitted his exit will have a very small impact, “but not no impact; some of my skills are rare”. The timing, he said, has nothing to do with the recent wave of viral resignations. He had picked the date long ago, ahead of a ten-day backpacking trip with friends on New Zealand’s Abel Tasman and Wangapeka tracks.He is not walking away from AI altogether. O’Callahan will keep maintaining the debugging tools rr and Pernosco, and plans to study how AI agents debug code. He wants his next project to be “unambiguously pro-human”.
