DeepMind Researcher Resigns Over Superintelligence Risks

Robert O'Callahan, a technical expert at Google DeepMind based in New Zealand, has resigned from his position due to growing concerns over the rapid pace of artificial intelligence development. O'Callahan, who specialized in creating chip design tools that made AI hardware both cheaper and faster, stated that he could no longer justify contributing to the technology's acceleration. He characterized the current velocity of industry progress as "far too high" and expressed deep reservations about the direction of frontier AI research.

In explaining his departure, O'Callahan warned that aiming to build artificial superintelligence in the near term is "inherently irresponsible." While he described the existential threat of superintelligence as real but uncertain, he highlighted several immediate societal dangers. These include cognitive surrender, AI-induced psychosis, widespread loneliness, extreme concentration of power, economic disruption, cybersecurity vulnerabilities, and a general lack of accountability among developers. Ultimately, O'Callahan noted he would wager that the collective harms of AI will outweigh its benefits, prompting his decision to pivot toward what he calls "unambiguously pro-human" work.

According to O'Callahan, his anxieties are not isolated within the company. He claimed that many of his former colleagues at Google DeepMind harbor similar apprehensions about the technology's trajectory but rarely voice them in public forums. His departure adds to a growing pattern of researchers leaving prominent AI labs over safety concerns. To support his views, O'Callahan referenced the essay "If Anyone Builds It" by Eliezer Yudkowsky and Nate Soares, figures associated with the effective altruism movement.

For AI practitioners and hardware engineers, O'Callahan's resignation underscores an escalating ethical dilemma within the industry. As technical experts find themselves building the very tools that accelerate AI capabilities, the pressure to balance rapid innovation with safety guardrails is intensifying. This tension is further complicated by political dynamics, such as a reported White House memo that characterizes the effective altruism movement as a fringe group and criticizes figures like Anthropic CEO Dario Amodei for fostering AI doomerism.