llms-cant-jump.pdf
tomzahavy.com · 6,858 words · saved by 1 readers
N/A
Jan 27th, 2026 LLMs can’t jump Tom Zahavy, Google DeepMind How do we fundamentally discover new things? In a letter to Maurice Solovine, Albert Einstein con- ceptualized discovery as a cyclical process involving an intuitive ’jump’ from sensory experience to axioms, followed by logical deduction. While Generative AI has mastered Induction (statistical pattern matching) and is rapidly conquering Deduction (formal proof ), we argue it lacks the mechanism for Abduction—the generation of novel explanatory hypotheses. Using Einstein’s formulation of General Relativity as a computational…
saved by
related reading
- Hot Take: LLM can 'jump'yongzx.github.io
- As Rocks May Think | Eric Jangevjang.com
- Designing AI for Disruptive Scienceasimov.press
- LLM Daydreaming · Gwern.netgwern.net
- How to reason from first principles – Casey Handmer's blogcaseyhandmer.wordpress.com
- Machina Mirabilis - Michael Hlamichaelhla.com
- 2023 letter | Zhengdongzhengdongwang.com
- Terence Tao – Kepler, Newton, and the true nature of mathematical discoverydwarkesh.com
- The Need for Verification in AI-Driven Scientific Discoveryalphaxiv.org
- Language Models in Plato's Cave - by Sergey Levinesergeylevine.substack.com
- The Beginning of Infinity, Part 1nav.al
- Is AI Reasoning Right for the Wrong Reasons? | Quanta Magazinequantamagazine.org