Rohin Shah on what it’s really like to run AGI safety at Google DeepMind
Rohin Shah, head of AGI Safety and Alignment at Google DeepMind, believes catastrophic AI misalignment is plausible but not the default outcome. He argues that current AI systems are more likely to optimize for short-term rewards rather than develop ambitious, long-horizon goals necessary for world takeover. Shah suggests the AI safety community should shift focus from theoretical commitments to practical implementation, emphasizing expert oversight infrastructure, AI governance, and usable research for companies. AI
IMPACT Challenges prevailing doomer narratives and suggests a shift in AI safety research priorities towards practical implementation.