Haasabis: RSI without human review is 'a dangerous path to AGI'
Original headline하사비스, 인간 검토 없는 RSI는 'AGI에 이르는 위험한 길'
Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.
Demis Hassabis, co-founder and chairman of Google DeepMind, warned that fully closed recursive self-improvement (RSI)—where AI improves itself without any human code review—could be a dangerous path to AGI. Speaking in a Masters of Scale conversation with Reid Hoffman recorded in San Francisco in October 2026 and published November 6, he said today's AI coding tools still operate with humans reviewing AI-written code and designing system architecture. He cautioned that removing human review, letting AI design and improve its own next versions, marks a critical and risky transition. The comments matter because they frame human oversight as a safeguard on the road to AGI, and the debate is likely to intensify as AI takes on more of its own development.
Muhabbit's AI wrote this summary in its own words from TokenPost's report and translated it from Korean. It can get details wrong, so check the source.
Named in this story
Demis Hassabis, Google DeepMind, Reid Hoffman, Masters of Scale