Will a large language model beat a super grandmaster at chess by EOY 2035?

Prediction market on manifold. This market will resolve if a chess super grandmaster is beaten in a match by a large language model before 2036. The point of this market is to predict if accessible LLMs will become smarter than expert humans. The chosen medium for this test is chess. Constraints: Human chess player must have a classical ELO of over 2700 Human does not need to be blindfolded. LLM must be generalised for everyday use, not trained on chess specifically. Chess match must be ‘taken seriously’: Match is streamed publicly. Match has sufficient anti-cheating measures for both sides. Match must include multiple chess games, each player must play white at least once. LLM must be available to the general public. LLM must not incur unreasonable costs during the match. LLM must not use any tools external to its trained machine learning parameters. I.e. no running code, no accessing a chess engine, no external internet access (similar restrictions to the human player). LLM must be input human moves via a natural language interface, can be text, audio, or visual. LLM is allowed to write as much as it wants for reasoning purposes. LLM output will not be restricted using structured outputs or any chess-specific schema. I.e move validity will be checked by a human before being applied to the board.

24h Volume: $30. Liquidity: $1,000. Resolves: 12/31/2035.