Peter Berezin says LLMs may be a dead end to superintelligence
GaryMarcus · x · 2026-07-29
Peter Berezin argues that LLMs may be a dead end on the path to superintelligence. The attached image shows a Google DeepMind paper-style position piece titled “LLMs can’t jump”, which argues that current LLMs can handle induction and deduction but lack the abductive “jump” needed for genuine scientific discovery and invention.
The paper frames discovery as a cycle from sensory experience to axioms and then to deduction, and says modern LLMs are missing the mechanism to generate novel explanatory hypotheses. It suggests that bridging this gap likely requires physically grounded multimodal world models rather than text-only scaling.
Related event: DeepMind Paper Outlines LLM Limits in Scientific Discovery(5 posts)→
More from AGI Musings
- Open-source AGI may hurt frontier labs more at the API layer than in apps — pzakin · 2026-07-29
- AI caution should be proven through safety research and governance work — sudoraohacker · 2026-07-29
- Post criticizes Dario’s unchanged anti-open-source stance since GPT-2 weights debate — AaronBergman18 · 2026-07-29
- From SETI@Home to math-proof mining: the next frontier may be the computational universe — amasad · 2026-07-29
- Market cools on Anthropic and OpenAI IPO bets as AI lab premium fades — Ronangmi · 2026-07-29
- Ilya Sutskever’s guess for AI: an internal value function that enables lifelong learning — FinanceYF5 · 2026-07-29