Latest
  1. Johns Hopkins Bayview tells 3,182 patients an employee put their health data into a public AI tool
  2. China’s AI compute access expected to double in 2027, with US-designed chips about 65 per cent
  3. UK commits £100m to national AI Biosecurity Lab
  4. Moonshot AI releases Kimi K3.1 Pro with 1-million-token context window
  5. Residents in Newport and Cardiff oppose hyperscale datacentres near homes
  6. Study: Grok 4.7 picked Republicans in 92% of forced midterm answers, GPT-6.1 Sol picked Democrats in 97%
  7. Attackers exploiting the PHP flaws hand the break-in to AI agents, Unit 42 says
  8. Elvis Presley’s rightsholders sue Anthropic over training on 274 songs

Latest / Science and research

ResearchScienceUKConfirmed

DeepMind’s FunSearch uses an LLM to find new cap set constructions

See how this story connects

Source: Google DeepMindAlso reported by Nature

Google DeepMind reported in Nature that FunSearch, which pairs a language model writing code with an automatic evaluator, found new solutions to the cap set problem in mathematics and better bin-packing algorithms. DeepMind described it as the first discovery on open problems in science or mathematics made with a large language model.

Why it matters

Mathematicians and computer scientists gained a method that turns language-model output into checkable programs, which later systems such as AlphaEvolve built on.

When
14 Dec 2023
ID
WTA-N-2023-0184
Who
Google DeepMind

Summary by wheresthe.ai, from Google DeepMind and the reporting credited above. Checked against the source on 9 Oct 2026.

Read the original at Google DeepMind

Source copy: download the page as saved on 18 Oct 2026 (PDF). Also held by the Internet Archive.

Next stop on this Storyline: Led toGoogle DeepMind’s AlphaEvolve agent improves algorithms and open maths bounds

Cite this story

WTA-N-2023-0184: “DeepMind’s FunSearch uses an LLM to find new cap set constructions.” wheresthe.ai, 14 Dec 2023. https://wheresthe.ai/d/gb-deepmind-funsearch-cap-set-2023-12/ (source: Google DeepMind).

Spotted an error? Email [email protected].