Schut Lisa, Tomašev Nenad, McGrath Thomas, Hassabis Demis, Paquet Ulrich, Kim Been
Oxford Applied and Theoretical Machine Learning Group, Department of Computer Science, University of Oxford, Oxford OX1 3QG, United Kingdom.
Google DeepMind, London N1C 4AG, United Kingdom.
Proc Natl Acad Sci U S A. 2025 Apr;122(13):e2406675122. doi: 10.1073/pnas.2406675122. Epub 2025 Mar 26.
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 concepts are at the frontier of human understanding. Despite the small scale, our result is a proof of concept demonstrating the possibility of leveraging knowledge from a highly capable AI system to advance the frontier of human knowledge; a development that could bear profound implications and shape how we interact with AI systems across many applications.
人工智能系统在各个领域都取得了超越人类的表现。如果能够利用这些高性能系统中编码的隐藏知识，人类的知识和表现就能得到提升。然而，这种内部知识很难提取。由于可能的内部表示空间巨大，寻找有意义的新概念知识就如同大海捞针。在此，我们介绍一种从AlphaZero中提取新国际象棋概念的方法，AlphaZero是一个通过自我对弈在无人类监督的情况下掌握国际象棋的人工智能系统。我们的方法使用凸优化从AlphaZero的内部表示中挖掘代表概念的向量，并基于可教性（该概念是否可转移到另一个智能体）和新颖性（该概念是否包含人类国际象棋游戏中不存在的信息）对概念进行筛选。这些步骤确保所发现的概念是有用且有意义的。对于所得的概念集，向专家展示原型（国际象棋谜题 - 解决方案对）以进行最终验证。在一项初步的人类研究中，对四位顶尖国际象棋特级大师（均为前世界冠军或现任世界冠军）解决概念原型棋局的能力进行了评估。所有特级大师在学习阶段后都有进步，这表明这些概念处于人类理解的前沿。尽管规模较小，但我们的结果是一个概念验证，证明了利用高性能人工智能系统的知识来推进人类知识前沿的可能性；这一发展可能会产生深远影响，并塑造我们在许多应用中与人工智能系统交互的方式。
