Chess is known for cold reasoning, yet the latest chess engines are shunning known techniques
When we talk about computers and chess, most people’s first memory will probably be Garry Kasparov losing to a computer in 1997. The computer in question was no ordinary PC – it was a supercomputer in every sense of the word. Search for the machine today, and the first images you will find will probably be the remnants of the machine that are now on display at the Computer History Museum.
Unfortunately, if you are interested in visiting the museum, it is closed until the Summer. Still, it can be found in the heart of California’s Silicon Valley – the home of many of today’s technology titan’s such as Apple, Facebook, Google, eBay and Adobe.
The machine on display looks like an ordinary mainframe – a huge cabinet of rack-mounted servers. The thing is, the machine on display is only HALF of the original machine – it took two of these enormous machines, over two meters tall and a little over a meter wide –to hold the 120 nodes of custom PS2C chips, each enhanced with 480 VLSI (Very Large Scale Integration) chess computing cores to beat Garry Kasparov that day.
An interesting side note is that these VLSI chips are a very early example of the Application Specific Integrated Circuit (ASIC) designs used today to mine Bitcoins and other cryptocurrencies.
Deep Blue: What Happened?
The original idea behind Deep Blue was to create a machine that could demonstrate a form of Artificial Intelligence. However, in reality, the computer was using brute force and data from millions of past games to counter the moves played by Kasparov. Analyzing the games with the latest chess engines such as Stockfish reveals that neither Kasparov nor Deep Blue was on their best form during the competition. Had Deep Blue not made a mistake in the second game – a move that shook Kasparov and led to him accusing IBM of cheating – he would surely have spotted that he had the opportunity to force a draw by perpetual check.
Garry had played Deep Blue the previous year and achieved a fairly decisive victory of 4-2. His loss in 1997 was slight – 3½–2½. Kasparov had won the first game, but the rest of his points came from the draws he forced in the next three games before he suffered a humiliating loss in the final game. Had Kasparov been on form, he could at the very least have achieved a draw in this contest
After Deep Blue
IBM continued to develop ever stronger chess machines after Deep Blue, though Kasparov declined a match against the newer iterations. Kasparov lost his world champion title to Vladimir Kramnik in 2000. However, Kasparov did continue to play chess professionally until 2005, when he won the prestigious Linares International Chess Championship for the ninth time. That contest was very closely fought, both with betting outfits such as at canadian bookmaker Unibet, and across the board, with Kramnik putting up a tough fight for the 9-time winner.
Vladimir Kramnik was more willing to get involved in Chess computing and eventually worked with the Google team who created the “true” artificial intelligence machines AlphaOne and AlphaGo. Together, they came up with various games that slightly altered the rules of chess, making the game much more difficult for a computer to play and less predictable for human players.
One example of this is Chess960, a game where the positions of the powerful pieces on the back rank of the board are randomized at the start of each game. Google’s DeepMind team were able to work with Kramnik to tap AlphaZero’s ability to learn a game from scratch and explore new variants more quickly than a human player ever could. “You don’t want to invest many months or years of your life trying to play something, only to realize that, ‘Oh, this just isn’t a beautiful game,’” said NenadTomašev, a DeepMind researcher who worked on the project.
So, Have Computers Reinvented Chess?
Not just yet. Kramnik, who retired from competitive chess last year, believes his beloved game has become less creative, partly due to computers. He blames their “soulless calculations”, which have produced a vast library of openings and defences that top-flight players now know off by heart. Some games are being played mostly from memory due to this – although the best players are spending their time learning how to counter these known techniques rather than working harder at memorizing known patterns.
If players are willing to accept some small changes to the rules, though – changes that computers have suggested would make the game less predictable – then computers will indeed have finally found a way to reinvent the modern game of chess.
