ABAB Pedia

Rishi Mehta

Overview

Contents8 sections
Key facts

Chess and the path to mathematics research

Mehta had worked in computer science and machine learning and played chess. Reading the AlphaZero paper and studying games the program had played convinced him to pursue similar work. When he joined DeepMind, Thomas Hubert’s AlphaZero team was working on mathematics. Mehta subsequently became one of the technical leads of AlphaProof. He met his favorite player, Hikaru NakamuraHikaru NakamuraHikaru Nakamura is a Japanese-born American chess grandmaster who at 15 became the youngest American grandmaster up to that time, has won the U.S. Chess Championship five times and the 2022 FIDE Fischer Random World Championship, and is a prominent chess streamer.Open the full entry, at a Paris blitz tournament in 2016. As a subscriber to Nakamura’s stream, he got to play him online in January 2020 and played a second game a few months later.⁠[3][1][4]

Machine-checked proofs of Olympiad problems

In 2024 the AlphaProof team and the team behind AlphaGeometry 2 tested the systems on the six problems from that year’s International Mathematical Olympiad. People translated the problems into formal language. AlphaProof solved two algebra problems and one in number theory. AlphaGeometry 2 solved the geometry problem, while the two combinatorics problems remained open. Mathematicians awarded the combined work 28 of 42 points, the silver-medal standard that year. The systems took up to three days on some problems, compared with two sessions of four and a half hours each for the human contestants. The teams subsequently published the research.⁠[1][2]

Interview with No Priors

On November 14, 2024, No PriorsNo PriorsNo Priors is an AI interview podcast and media brand founded and hosted by investors Sarah Guo and Elad Gil, with episodes beginning in February 2023. It covers research, startups, infrastructure and business applications through guest interviews, host discussions and retrospectives distributed on audio platforms and YouTube.Open the full entry released an interview with Rishi Mehta. He explained how chess led him to work on machine-generated proofs and described the appeal of pursuing a mathematical question for its answer even without an obvious application.⁠[3]

Mehta said his interest in chess made AlphaZero’s games particularly striking. They drew him to similar research, and he moved into mathematical proofs after joining DeepMind.⁠[3]

Work, travel and writing

Before DeepMind, Mehta used deep learning to improve GoogleGoogleGoogle is an American technology company founded by Larry Page and Sergey Brin and incorporated in 1998. It became an Alphabet subsidiary in 2015. Its businesses include internet search, advertising, software platforms, cloud computing, and artificial intelligence.Open the full entry’s search ranking. At DeepMind he also worked on post-training Gemini. He later joined AnthropicAnthropicAnthropic is an artificial intelligence research and product company founded in 2021 that develops the Claude models, assistant, and tools such as Claude Code, and operates as a public benefit corporation. Dario Amodei and Daniela Amodei are its co-founders, and its research covers model safety, interpretability, and steerability.Open the full entry, where he worked on Claude’s coding capabilities from Opus 4 onward, and lived in San Francisco. In February 2025 he wrote that he had moved nine times in eight and a half years since leaving college, including three moves between countries. On three other occasions he had lived out of a bag for two or three months. In March he wrote about spending a morning with a Hadza group in Tanzania. He also built Chorus, a browser extension for comparing language model responses.⁠[5][6]

Advice Shared on No Priors

He encouraged mathematicians to consider formal mathematics as a way to collaborate on checkable proofs, while noting that Lean users were still a minority. (35:04)⁠[3]

Related Organizations, Websites and Public Accounts

Google DeepMind: Co-lead of the AlphaProof project. Official website: https://deepmind.google/ (opens in a new window).⁠[1]

Anthropic: Researcher. Official website: https://www.anthropic.com/ (opens in a new window).⁠[5]

Sources6 sources

  1. AI achieves silver-medal standard solving International Mathematical Olympiad problems (opens in a new window)deepmind.google↑
  2. Olympiad-level formal mathematical reasoning with reinforcement learning (opens in a new window)research.google↑
  3. No Priors Ep. 90 | With Google's DeepMind's AlphaProof Team (opens in a new window)youtube.com↑
  4. My chess games against Hikaru (opens in a new window)rishimehta.xyz↑
  5. About (opens in a new window)rishimehta.xyz↑
  6. Rishi Mehta: Blog (opens in a new window)rishimehta.xyz↑