Oxford paper argues LLMs can't truly innovate because they imitate data, not build theory
GregCook2011 · x · 2026-09-19
Oxford Saïd Business School researchers Teppo Felin and Matthias Holweg published "Theory Is All You Need," pushing back on the AI-will-replace-everyone narrative.
Key claims:
- LLMs predict the next token from existing data, making them structurally backward-looking and imitative;
- A model cannot hold a belief the training data says is wrong — yet every major breakthrough in history began with exactly such a belief;
- The paper's central example: the Wright Brothers flew in 1903, nine weeks after expert consensus said human flight was one to ten million years away — they had a theory, not better data.
The authors conclude LLMs are unlikely to produce genuinely novel inventions.
More from AGI Musings
- eigenrobot shrugs at AI 'independent proof' plagiarism scenario raising ethics debate — eigenrobot · 2026-09-19
- APA warns: kids and teens are turning to AI for friendship, advice, and emotional support — AnnaCiaunica · 2026-09-19
- Falsifiable test proposed for what triggers model pain directions — EigenGender · 2026-09-19
- "I'm a single-issue voter on AI": the anti-frontier-lab stance summed up in one post — AIandDesign · 2026-09-19
- Debate: adding opaque recurrences to chain-of-thought makes monitorability dramatically worse — panickssery · 2026-09-19
- Harari: as AI automates finance at machine speed, accountability cannot be automated — Olivier__OG · 2026-09-19