When Moore's Law Killed Chess

How Strategy Games Redefined Intelligence in AI

Abstract

The history of AI is a history of games. This piece charts a centennial metamorphosis of machine intelligence in strategy games, from El Ajedrecista in 1912 to Quantum Go in 2020. The purpose of this paper is to highlight modern breakthroughs and disruptions in the way scientists understood machine intelligence, like reinforcement learning and quantum computing. For hundreds of years, chess was associated with intellectual ability, and, in the 20th Century, became the cornerstone of AI research. Now, in the 21st Century, researchers have expanded their work to include the games of Go and shogi, and even revisited chess with new algorithmic approaches previously unattainable. Traditional game-theoretical approaches to computational decision-making have hit a ceiling due to hardware limitations as there are now more permutations of choices and positions than atoms in the universe, way more than a computer could handle. According to the 2019 Stanford AI Index, AI’s heavy computational requirement outpaces Moore’s Law, doubling every three months rather than two years. This challenged scientists’ ability to build intellectual machines. Thus, they grappled with what it meant for a machine to be intelligent and what people could accomplish with intelligent machines.