Chollet: Don't declare AGI until AI can truly invent
fchollet · x · 2026-09-08
François Chollet argues the narrative for building AGI has always been about invention — curing cancer, solving fusion. So no one should 'declare AGI' until AI is capable of conceptual breakthroughs, novel insights, and real-world new technology. An AGI system, he says, must produce more than what you put into it.
Related event: Chollet: Don't declare AGI until AI can truly invent(3 posts)→
More from AGI Musings
- Gary Marcus responds to 'you can't claim AI is both dangerous and useless' critique — GaryMarcus · 2026-09-08
- Gary Marcus rips Jensen Huang for claiming AGI arrived 'with no evidence and no definitions' — GaryMarcus · 2026-09-08
- Agents automate the boring high-leverage work — and veterans benefit most — sloppenheimer · 2026-09-08
- SymphonyAI employee says leadership demands 2–3x productivity amid AI struggles — Amazing_Respond6631 · 2026-09-08
- AI 2027 and AI as Normal Technology Authors Co-Author Essay on Their Common Ground — AndyMasley · 2026-09-08
- SymphonyAI insider: 'vertical AI' may be a transition category, not a moat — Amazing_Respond6631 · 2026-09-08