Robert O’Callahan has left his role as a technical expert at Google DeepMind in New Zealand after concluding that accelerating artificial intelligence development is inherently irresponsible. O’Callahan, who previously contributed to chip design tools that reduce costs and increase processing speed, states the current rate of change is far too high to justify his continued involvement. He argues that while the risk from superintelligent AI is real, aiming for it in the near future creates unacceptable dangers. These include cognitive surrender, AI-induced psychosis, loneliness, concentrated power, economic disruption, cybersecurity risks, and a lack of accountability. O’Callahan believes the potential harms outweigh the benefits and now seeks work that is unambiguously pro-human. He references arguments by Eliezer Yudkowsky and Nate Soares without endorsing their claim that superintelligent AI inevitably brings catastrophe. Many colleagues share his concerns but rarely speak publicly, while the White House has recently cast the effective altruism movement as a cult-like fringe group.
This departure highlights a growing internal disagreement within major technology firms regarding safety timelines and ethical boundaries. It suggests that individual engineers are increasingly willing to challenge corporate priorities when those priorities conflict with their personal risk assessments. The situation underscores the difficulty of aligning rapid technical progress with societal safeguards.
* O’Callahan worked on hardware tools that made AI models cheaper and faster to train.
* The White House memo criticises the effective altruism movement for favouring charity over redistribution.
* Many DeepMind staff hold similar views but remain silent to protect their employment.
