Gary Marcus says a 900-ELO chess bot is nowhere near AGI
GaryMarcus · x · 2026-07-29
Gary Marcus pushes back on claims that Fable is “close to AGI,” arguing that a system that can only play chess at around 900 ELO is nowhere near human-level general intelligence. He adds that a young Garry Kasparov would likely have beaten the best commercial LLMs at chess, underscoring how weak such comparisons can be.
Related event: Gary Marcus Refutes LLM AGI Claims Using Chess Skills(2 posts)→
More from AGI Musings
- Six weeks after the last frontier model launch, AI still looks brittle — bindureddy · 2026-07-29
- Miles Brundage says AI safety needs concrete models of the worst-case failures — Miles_Brundage · 2026-07-29
- AI power could make ideology secondary to survival, this reply argues — zetalyrae · 2026-07-29
- New RSIBench-Data benchmark shows agents improve model data pipelines, but often overshoot — imjustnewatai · 2026-07-29
- Should the AI race pause for safety or because models are already powerful enough? — mallow610 · 2026-07-29
- Why one AI bull says the boom still has room to run: TAM, RSI, agents and HBM — johncoogan · 2026-07-29