Fired OpenAI Researcher's Startling Essay AI Models Could Surpass Humans by 2027, Predicts 'Intelligence Explosion'

  • 12 days ago
Leopold Aschenbrenner, a former researcher at OpenAI, published a 165-page essay discussing his predictions about the future of AI. Aschenbrenner worked on OpenAI's superalignment team to help mitigate AI risks before being fired in April 2024. Aschenbrenner argues that AI progress is accelerating rapidly and models could reach human-level capabilities as early as 2027. He predicts an "intelligence explosion" could then occur where AI surpasses human intelligence quickly after reaching general intelligence. Managing superintelligent AI will be crucial to prevent potentially catastrophic outcomes if it is not properly aligned with human values.