Sam Altman: AI already speeds up OpenAI's research, but a self-improving button? 'Definitely not'

ChrisGPT · x · 2026-11-17

Asked whether RSI (recursive self-improvement) is already happening, Sam Altman said AI is genuinely helping OpenAI develop models faster — something observable since the GPT-5.x era — but denied that anyone can 'push a button' and have AI rapidly improve itself without human oversight: 'definitely not.' He compares defining RSI to defining AGI, noting neither has a clean dividing line.

Original post →

More from AGI Musings

AGI Musings channel →