Chess and the Search for AI

Garry Kasparov’s latest book asks important questions about Artificial Intelligence, using his own battle with Deep Blue as a guide.

Garry Kasparov’s book, Deep Thinking: Where Machine Intelligence Ends and Human Creativity Begins, explores his own journey that is intertwined with humanity’s project to make chess-playing computers.

Kasparov and his co-author Mig Greengard (also Kasparov’s spokesman) succeed in making both machine intelligence and chess intelligible for the lay reader. The book is full of insights and historical context on both the development of modern chess as well as machine learning.

Deep Thinking articulates the key insight of modern machine intelligence succinctly: the difference between general intelligence and the machine’s ability to surpass humans in specific skills. The two fallacies are:”The only way a machine will ever be able to do X is if it reaches a level of general intelligence close to a human’s”; and “If we can make a machine that can do X as well as a human, we will have figured something profound about the nature of intelligence.” As Kasparov argues, time and again attempts to make machines that think like humans have failed, while machines that prioritize results over method have succeeded. We don’t need to ape nature, even if we are inspired by it. For example, airplanes don’t flap wings, and the wheel doesn’t exist in nature.

Chess became the drosophila that researchers used to explore what it takes to create an artificial mind. The book introduces you to Claude Shannon, the father of Information theory, who wrote the first chess-playing program in 1950. Shannon was brilliantly prescient when he said, “Chess is generally considered to require ‘thinking’ for skillful play; a solution to this problem will force us either to admit the possibility of mechanized thinking or to further restrict our concept of ‘thinking.'” Shannon also figured the two approaches to playing chess: a) Brute force – just calculate as many possibilities as the computer can and then evaluate which move is the best. b) Figure out a good move through an insight.

Over years, brute force has won. Hands down.

Kasparov’s own eagerness to participate and explore how far a chess-playing computer can go shines through the book. His interest in this field started in the 80s itself. It was clear to the researchers by the late 1980s that Chess held no special insights in unlocking the mysteries of the machine intelligence. In 1990, Ken Thomson even declared the Chinese strategy game of Go as the new drosophila for AI (Google’s alphaGo mounted this summit early this year). But we digress; this book is largely the story of how IBM’s machine Deep Blue beat Gary Kasparov in 1997. Even though creating a chess machine that beat a World Champion was not going to reveal anything more in the cause of artificial intelligence, the public perception of Chess as an intellectual pursuit yielded it the high drama that IBM, and perhaps Kasparov, could not shy away from.

This battle between Deep Blue and Kasparov is roughly half the book. Although one can understand Kasparov’s need to get his side of the story straight, the ball-by-ball nature of the account (trust an Indian review to bring cricket into chess) doesn’t add to the promise of the book. Kasparov does manage to convey why he accused IBM of cheating or being unfair. The psychological nature of machine-vs-man sporting battles also gets highlighted well. In games such as Chess, you are playing the opponent as much as the game. However, for the machine, the opponent is irrelevant, and it only has the cold pursuit of the best move given a board position. Kasparov often attributes greater intelligence to Deep Blue during the game than subsequent analysis shows. Instead of obsessing over the game, he is obsessing over the machine.

Leaving aside the fact that the long middle portion of the book distracts from the fundamental questions, it is a racy and enjoyable read. One key insight does tie to the conclusion of the book that Kasparov argues for: humans plus machines will be a more formidable combination than any of them alone. In fact, Kasparov coins a law that Humans+Machines+Process is the key, and weak humans+machines+strong process will prevail over strong humans+machines+weak process. This was something he learnt from subsequent chess tournaments that he organized where players could team up with machines. The lesson from the Deep Blue saga was that the match wasn’t between the machine and Kasparov but between IBM and Kasparov. IBM used a machine, a team of people and a process (of questionable ethics) to outsmart the world champion.

Kasparov raises important points about how current AI approaches can’t figure out things that humans are good at: context and purpose. He doesn’t follow through with an exploration of answers, though. Is motivation the preserve of biological intelligence that has evolved in the hyper-competitive world of natural selection? Or is it a matter of deciphering the mathematical code behind motivation to program a machine? He simply quotes researcher Andrew Ng to kick the can into the future – Worrying about super-intelligent and evil AI today is like worrying about the problem of overcrowding on Mars.

As Kasparov’s own saga shows, we don’t have to worry about evil AI in the near future, but the prospect of evil humans using AI to further their goals is more likely. This conclusion is too pessimistic for Kasparov’s taste though. He himself ends on a more optimistic note, asserting that technology will free us to be more creative, and that the future lies in humans using machines to achieve their full potential. Read the book. Kasparov is very persuasive and even if he doesn’t leave you optimistic, he will leave you wiser.