{"slug": "top-mathematicians-say-llms-are-strong-calculators-but-poor-creative-thinkers", "title": "Top mathematicians say LLMs are strong calculators but poor creative thinkers", "summary": "Mathematicians Timothy Gowers and Peter Sarnak say large language models (LLMs) are strong at calculation but lack the intuition for creative mathematical breakthroughs, according to separate essays published in August 2026. DeepMind researcher Tom Zahavy's paper \"LLMs Can't Jump\" similarly argues that LLMs cannot invent new foundational assumptions, a process he calls \"manipulative abduction,\" and suggests world models as a potential path forward.", "body_md": "# Top mathematicians say LLMs are strong calculators but poor creative thinkers\n\n**LLMs can't jump, Part II.** Mathematicians Timothy Gowers and Peter Sarnak credit large language models with serious math skills but see hard limits for genuinely new ideas. [Gowers argues](https://gowers.wordpress.com/2026/08/12/what-sort-of-maths-are-llms-good-at/) current models are good at combining known methods and trying many search paths but lack the intuition to pick the few productive routes in a vast search space. [Sarnak agrees](https://www.ams.org/journals/notices/202607/noti3373/noti3373.html): AI can derive results from existing theory but fails to develop the abstractions that underpin major proofs when starting from an elementary question.\n\nDeepMind researcher Tom Zahavy reached a similar conclusion. In his [paper \"LLMs Can't Jump,\"](https://the-decoder.com/language-models-cant-spark-scientific-revolutions-but-world-models-might/) he pins the bottleneck on \"manipulative abduction,\" the ability to invent new foundational assumptions with no linguistic precedent. World models could offer a way forward. These assessments feed into a broader debate about [whether LLMs are actually becoming more versatile](https://the-decoder.com/ex-openai-researcher-bets-100-billion-will-flow-into-training-data-because-scaling-alone-wont-cut-it/) or \"just\" getting better at benchmarks and familiar problem spaces.\n\n```\nAI News Without the Hype – Curated by Humans\n\n\t\t\t\t\tSubscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive \"AI Radar\" frontier report six times a year, full archive access, and access to our comment section.\t\t\t\t\n\n\t\t\t\t\tSubscribe now\n```\n\n", "url": "https://wpnews.pro/news/top-mathematicians-say-llms-are-strong-calculators-but-poor-creative-thinkers", "canonical_source": "https://the-decoder.com/top-mathematicians-say-llms-are-strong-calculators-but-poor-creative-thinkers/", "published_at": "2026-08-16 15:31:21+00:00", "updated_at": "2026-08-16 15:42:13.702964+00:00", "lang": "en", "topics": ["large-language-models", "artificial-intelligence", "ai-research"], "entities": ["Timothy Gowers", "Peter Sarnak", "DeepMind", "Tom Zahavy", "LLMs Can't Jump"], "alternates": {"html": "https://wpnews.pro/news/top-mathematicians-say-llms-are-strong-calculators-but-poor-creative-thinkers", "markdown": "https://wpnews.pro/news/top-mathematicians-say-llms-are-strong-calculators-but-poor-creative-thinkers.md", "text": "https://wpnews.pro/news/top-mathematicians-say-llms-are-strong-calculators-but-poor-creative-thinkers.txt", "jsonld": "https://wpnews.pro/news/top-mathematicians-say-llms-are-strong-calculators-but-poor-creative-thinkers.jsonld"}}