DeepMind's O'Callahan quits, calls near-term ASI irresponsible

O'Callahan's exit is the latest in a series of Google researcher departures over AI risks, and his own framing is narrower than the loudest version of the debate: he calls the risk from superintelligent AI "real but uncertain" rather than certain, and he declines to endorse the argument in "If Anyone Builds It" that catastrophe is inevitable. What makes his case distinctive is what he worked on. His chip design tools helped make AI cheaper and faster, and it is precisely that contribution he says he can no longer justify, which puts the practical efficiency work behind modern AI models at the center of his reasoning rather than abstract speculation.

He also points to a gap between private and public sentiment inside the lab. Many DeepMind colleagues share his concerns but rarely speak publicly, he says, which suggests the visible resignations may understate how widespread that unease is. Beyond the lab, the article places his stance in a broader argument about effective altruism, the movement tied to the book's authors. A White House memo, per Axios, casts that movement as a cult-like fringe group and attacks Anthropic CEO Dario Amodei as the face of AI "doomerism."

Whether this resignation changes anything likely hinges on whether others follow him into public dissent, and on how labs respond to staff who want what he calls "unambiguously pro-human" work. For the companies building frontier systems, the practical question is retention: losing people with hands-on knowledge of how to make AI cheaper and faster is a different kind of cost than losing headline researchers.