Byzantine Reality

Searching for Byzantine failures in the world around us

Behind Deep Blue

While at the Computer History Museum a while ago for the Big Data Workshop, I spent slightly too long in the gift shop nerding out and on impulse, picked up Behind Deep Blue: Building the Computer that Defeated the World Chess Champion. My reasons were two-fold and pretty straightforward: I like computers and I like chess, so why not actually learn about the computer that changed the world more than a decade ago? But my expectations going in were surprisingly far off, as it’s not so much a book about computers and chess as it is a book about people who happen to use computers and play chess (and use computers to play chess of course). Yet it is the first book that I read cover-to-cover in a single sitting and simply could not put down.

This book is not for everyone. If you saw the title of the book and weren’t instantly intrigued, then the book probably isn’t for you. It follows the exploits of the author, Feng-Hsiung Hsu, as he eventually gains interest in building a chess computer that can defeat the World Chess Champion. For the uninitiated, this was seen as a monumental task and many people simply assumed that a computer could never play chess better than the masters. There’s a bit of hubris involved in the reasoning behind this, and it tends to boil down to “it could play chess ok, but lacks the ‘soul’ or ‘humanity’ to really play chess at the highest levels”. With the joy of hindsight, we can now see that this was similar to the Titanic, with people saying “this boat cannot sink” before the maiden voyage.

So quite a bit of time had been spent by groups worldwide trying to build a chess computer with this goal ultimately in mind, but a lot of them could play chess OK (and no better). The book covers the three big projects on chess computer development at Carnegie Mellon University and IBM that Hsu was involved in: ChipTest, Deep Thought, and Deep Blue. If you’re not an electrical engineer or computer engineer, some of the finer points on the hardware will be lost on you (which is why the book is not for all), but most will be able to appreciate to some extent the drive that formed in Hsu over the 12 years of his struggle to be the best.

(Deep Blue, photo taken at the Computer History Museum)

The book actually has a very sad ending, which I never saw coming. As the Titanic reference pointed out, you likely know going into this story that by the end, Deep Blue will defeat Garry Kasparov, the World Chess Champion. But the sad ending comes on, strangely enough, right when Deep Blue wins. This win comes in the 1997 rematch, and in the 1996 match, Kasparov wins without too much trouble. In the 1996 match, everyone is happy for Kasparov and cheers him on as something akin to the “protector of humanity from the machines”, and in the book it is explicitly said that he felt quite a pressure to be this protector. The victory is seen as a triumph of man, so when the 1997 rematch occurs and Deep Blue wins, nobody is really happy for the Deep Blue team. His three-man team has put in 30 combined years of their lives into this project and it is seen as a failure for humanity, while all along Hsu and Company have been trying to tell people that this really is not a failure. In their eyes, this is the “triumph of man as a toolmaker over man as a performer” (the quote might not be exactly correct but is pretty close). Furthermore, Kasparov’s completely unsportsmanlike behavior during the rematch and the drama he caused afterwards makes the ending even more sad. He accuses the team of cheating, of making moves that the computer wouldn’t have made, and demands to see Deep Blue’s logs during the match (which is equivalent to reading its mind since he would then know all the moves it was going to make). I’m sure if accusations like those were leveled at him he would have simply stormed off instead of being surprisingly mature as the Deep Blue team was.

(Kasparov v. Deep Blue, photo taken at the Computer History Museum)

This brings us to another interesting point about the book. In some sections, adversarial characters sometimes act exceedingly irrationally, and Hsu spends a lot of time slowing down these moments to try to help us understand why (for example) Kasparov would do some of these outlandish things. He says it’s still not acceptable, but repeatedly tries to at least give a well-reasoned explanation of his adversaries’ disrespectful actions. This allows the book to stay personal and told through his viewpoint but removes a lot of the bias that you could otherwise see in it.

It ends on another sad note after that, to make things even more depressing. With their goal completed, the team disbands and go their separate ways, to do things completely unrelated to chess (not the sad part). However, Deep Blue is taken out of commission, as the resulting media firestorm by Kasparov convinces IBM that spending more money on chess computers is no longer necessary (two reasons here: one, they already beat him, and two, Kasparov’s demands were too extravagant for IBM to be able to afford both him and the chess computer project, causing them to abandon both). A much slower version of Deep Blue is left online for IBM employees to play with (not sure if it’s still up anymore) but is a shadow of the true Deep Blue.

Kasparov goes on to say that surely he could not lose again, and Hsu leaves IBM and takes the rights to Deep Blue with him, seemingly with a rematch in mind. Kasparov backs down, drama ensues, and Hsu eventually becomes disinterested with chess computers. Kasparov then says that he definitely won’t lose to a computer now, but was unable to defeat Deep Junior and X3D Fritz in 2003 (they tied in both occasions). Hsu claims that Deep Junior and X3D Fritz were far inferior to Deep Blue, as Deep Blue had a number of critical advances over Deep Junior and X3D Fritz despite being much older (if you want the technical details of why, go buy the book). I buy the arguments made, and it’s a shame that the best chess computer we even had doesn’t exist anymore. I can see why, however, and why there’s not much interest in resurrecting the project – again, read the book for details I’ve skipped on here – and it’s one of those “sad but true” kind of things.

As I said in the beginning, I couldn’t put it down even though I knew the ending going in, which for me, makes this book a keeper. If you’re into computers or chess, put down whatever you’re doing and go get_Behind Deep Blue_. It’s that good.