Events

Chess has long been a proving ground for AI and statistical reasoning. This talk takes a Bayesian look at two recent flashpoints in elite play. The centerpiece is joint work with Shiva Maharaj (Chess Ed) and Vadim Sokolov (George Mason) on the 2023 Kramnik–Nakamura controversy, in which former world champion Vladimir Kramnik publicly questioned Hikaru Nakamura’s 45.5 out of 46 streak in 3+0 online blitz on chess.com. Combining Anand’s prior on the prevalence of online cheating with the streak evidence, we compute a posterior of roughly 99.6% that Nakamura did not cheat. The case study illustrates two classic fallacies — the Prosecutor’s Fallacy on Kramnik’s side, and a misuse of cherry-picking that violates the likelihood principle on Nakamura’s side — and connects to the broader literature on fraud detection and streaks in sports. I will then survey related work with the same group: a Brownian-motion model for the probability that Magnus Carlsen reaches an Elo of 2900 and its implications for the K-factor; a neural-network valuation of (piece, square) combinations; and a comparison of Stockfish and Leela Chess Zero as competing paradigms — handcrafted search versus deep reinforcement learning — through Plaskett’s endgame study.

tbc

Overview

Modern neural networks operate at unprecedented scales across model size, data and compute. A central research problem is to understand how their performance scales with these factors, which guides how networks can be trained optimally at scale. In recent years, empirical heuristics for scaling have arguably driven much of the success of Large Language Models (LLMs). Theoretical work on scaling laws has also seen much fruitful progress, shedding light on empirical phenomena such as model collapse, emergence and training stability, while providing concrete practical insights on techniques such as hyperparameter tuning.

This three-day workshop will bring together researchers working at the frontiers of theoretical scaling laws to share their insights about the field. The workshop will be the first of its kind in the UK, inspired by successes of similar workshops in the US and Europe.

• The first half of the workshop consists of introductory tutorials, with the aim of equipping attendees with basic tools for framing and understanding problems in this field;
• The second half will feature talks on latest research advances.

The aim is for researchers across academia and industry to learn about and participate in this active field of research, which has seen many fruitful empirical outcomes.