Bridging the human-AI knowledge gap through concept discovery and transfer in AlphaZero

• 成果类型: Article
• 署名作者: Schut, Lisa; Tomasev, Nenad; McGrath, Thomas; Hassabis, Demis; Paquet, Ulrich; Kim, Been
• 署名单位: University of Oxford; Alphabet Inc.; DeepMind; Google Incorporated; Alphabet Inc.; DeepMind
• 刊物名称: PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
• ISSN/ISSBN: 0027-9453
• DOI: 10.1073/pnas.2406675122
• 发表日期: 2025-04-01
• 关键词: shogi; chess; go

摘要

AI systems have attained superhuman performance across various domains. If the hidden knowledge encoded in these highly capable systems can be leveraged, human knowledge and performance can be advanced. Yet, this internal knowledge is difficult to extract. Due to the vast space of possible internal representations, searching for meaningful new conceptual knowledge can be like finding a needle in a haystack. Here, we introduce a method that extracts new chess concepts from AlphaZero, an AI system that mastered chess via self-play without human supervision. Our method excavates vectors that represent concepts from AlphaZero's internal representations using convex optimization, and filters the concepts based on teachability (whether the concept is transferable to another AI agent) and novelty (whether the concept contains information not present in human chess games). These steps ensure that the discovered concepts are useful and meaningful. For the resulting set of concepts, prototypes (chess puzzle-solution pairs) are presented to experts for final validation. In a preliminary human study, four top chess grandmasters (all former or current world chess champions) were evaluated on their ability to solve concept prototype positions. All grandmasters showed improvement after the learning phase, suggesting that the is a proof of concept demonstrating the possibility of leveraging knowledge from a across many applications.