In the afterword, Kasparov quotes Joseph Weizenbaum: machines can decide, but they do not choose. Why does a machine do what it does? Every mechanized decision can be traced back, step by step, 1 by 0 by 1, to an earlier branch in its code. Eventually this reaches the inevitable conclusion of “Because you told me to.”
This is one of the most profound articulation of AI that I’ve read. This book is a brisk read into the history of computers in chess culminating in Kasparov’s retelling of the historic Deep Blue versus Kasparov match. Much of this I was impervious to, and of interest to me is how tightly computing history intertwines with chess. Many of the computing pioneers used chess as a tool to probe, explore and discuss computational intelligence.
The recount of the climax was absorbing and entertaining. Perhaps more salient than the arc of technology advancements that led to the showdown was the myriad of human drama and intrigue that underscored the entire spectacle. After all, the stakes are high and emotions strung tight. It was revealing that Kasparov studied the weakness of chess computers and played to a style that exploits its weaknesses. Despite a few jabs here and there, I thought the author provided an objective treatment of the events; not being a chess aficionado however, I’m none the wiser.
Early on, Kasparov introduced Claude Shannon’s conceptualization of “Type A” and “Type B” search techniques. Type A being “brute force” and Type B, “intelligent search”, which really means search guided by heuristics. In the seminal match with Deep Blue, the chess computer employed a combination of Type A search and specific strategic overrides that targeted Kasparov’s mental game and playing style to score the victory. It was unprecedented, of course, but these days the existence of plausible real-time brute forcing and endgame tablebases (which are precomputed perfect play endgame moves) mean that humans are vastly outmatched. After all, we cannot have perfect recall no matter how much we try.
Again, in the afterword Kasparov talks about the emergence and superiority demonstrated by AlphaGo and AlphaZero, which are specific forms of reinforcement learning algorithms used by Google to humiliate humans in games. He thinks that the way these algorithms create rules from scratch may be more akin and the first instance of what Shannon postulate as the third way (or Type C), which is search using principles and context inferred from the game itself. And as always, he ends in an optimistic note.
The chess and AI relationship goes deep and way back and it has been interesting to read about progress in technology and the game and how the affect each other. The book, while a breezy read, is mostly a lengthy recount of these developments and, to my chagrin, does not have too much to say about the more recent developments.