World chess champion Garry Kasparov provides a profound, revealing diary of his competitions against the best chess computers in the world, and how this offers metaphors for future human/machine interaction.
In a famous match in 1996 and an infamous rematch in 1997, world chess champion Garry Kasparov played against Deep Blue, the most advanced chess-playing machine of its time. Machines “think” differently than humans, Kasparov emphasizes, because machines use deductive logic, not the inductive logic of humans. Kasparov details the history of machine versus human in the world of Grandmaster chess and discusses human-machine collaborations that aid decision-making.
Since his retirement in 2005, Kasparov examined how chess informs decision-making processes. Facing down Fritz 3, Deep Blue and Deep Junior inspired Kasparov to ask: If computers can play chess at Grandmaster levels, what else can they do? Computers can make deductive calculations at extraordinary speeds, but cannot match human intelligence in inductive reasoning.
Battling each new generation of chess machines meant participating in a hallowed scientific quest, sitting at the nexus of human and machine cognition, and holding up the banner for mankind.Garry Kasparov
Kasparov believes the “transfer of labor from humans to our inventions” is the essence – and marked the dawn – of civilization.
“Type A” Brute Force
So far, no machine emulates how humans think, but machines can crunch data quickly to achieve results that no human could achieve.
In order to use chess as a way to better understand what computers and humans are good at and what they struggle with and why, the moves matter more than the results.Garry Kasparov
In his 1949 book, Programming a Computer to Play Chess, mathematician Claude Shannon explained that chess suits computers because chess expresses a sharply defined problem with straightforward moves and goals. Chess provides the parameters for a discussion about whether a computer can “think” or if it challenges humans’ concept of thinking.
Shannon devised “Type A” (brute force) and “Type B” (intelligent) search techniques. Type A crunches millions of possible moves. Type B requires a more holistic, more strategic approach, which humans perform far better than machines can perform.
Computers can’t generate questions – a human deductive skill. In chess, humans may give a machine the rules and allow it to figure out the game independently. IBM, however, prioritized creating a machine that would beat Kasparov over creating one that could think like a human.
Chess Machines vs. Grandmasters (GMs)
Today, machines may achieve 90% accuracy; they must reach 99.99% before humans can trust them with tasks such as driving a car.
A chess machine that thinks like a human and loses to the world champion isn’t going to make the news. And when a chess machine beats the world champion, nobody cares how it thinks.Garry Kasparov
When Kasparov encountered Deep Thought – Deep Blue’s predecessor – in 1989, he understood the machine had gigabytes of data available and parameters set by Grandmasters. Deep Thought didn’t possess the human capacity to sacrifice material – chess pieces – in the middlegame to advance its endgame. Deep Thought could only see three moves deep and was taught to protect its material at all costs.
When IBM offered Kasparov a match with Deep Blue in 1996, he asked for a winner-take-all prize structure, confident he would win. But Deep Blue had not played publicly, so Kasparov knew nothing of its capacities.
Kasparov lost the first game, but learned how this machine “thought” about chess. In the next game, Kasparov forced Deep Blue into a defensive position and wrote it down, ending the game in 73 moves. In the third game, however, Deep Blue improved at defense, and gained a draw. In the fourth game, Kasparov offered an early draw, seeing no viable way for either side to win. The IBM team did not accept. Armed with new understanding, Kasparov outmaneuvered Deep Blue to win the match, 4-2.
The high-profile match elevated IBM’s status, drove up its stock price and inspired IBM’s developers to work on a machine that would unseat the World Champion only 450 days later.
IBM upgraded Deep Blue at the cost of millions of dollars. Kasparov confesses he assumed – wrongly – that he won the first match due to his superior skills, not the machine’s poor play. The new version calculated moves more quickly and was smarter.
Kasparov realized too late that IBM accepted the rematch to beat Kasparov by any means. IBM refused to share games Deep Blue had played – with the help of Grandmasters – to prepare.
Of the many attempts to make me, and humanity, feel better about my loss to Deep Blue, the only effective one was that it was also a win for humans, since humans built the machine.Garry Kasparov
Kasparov competed against a machine that never tired, never second-guessed itself and had the best players in the world tweaking its game. When the machine crashed, Kasparov never knew if the crashes disguised human intervention. He beat Deep Blue in game one, thanks to Deep Blue’s attachment to material and one inexplicable move that weakened its position.
Machines don’t understand gamesmanship, or how to game their opponent’s psychology, but their programmers might understand both. IBM built pauses in Deep Blue’s game to “unsettle” Kasparov. In game two, Kasparov played a nuanced game with the Ruy Lopez opening. Deep Blue surprised him by expertly breaking down his defenses. Kasparov capitulated, and discovered he had resigned a game that should have ended in a draw – something he had never done before. He lost the match to Deep Blue, 3.5-2.5 (draws count as half wins to each player). To understand Kasparov’s extraordinary abilities, consider that a human chess player is six times more likely to climb Mt. Everest than he or she is likely to beat Garry Kasparov.
It’s wonderful if we can teach machines to think like we do, but why settle for thinking like a human if you can be a god?Garry Kasparov
Kasparov discovered that IBM cheated by, for example, having a Russian-speaking security guard eavesdrop on Kasparov’s strategy conversations and then tweaking Deep Blue to combat his plans.
Grabbed a Rock
Kasparov, unsurprisingly, is a careful, didactic, linear thinker who builds his arguments – and his paragraphs – idea by idea, brick by brick. He records his experiences in a dispassionate yet revealing voice and, given the ego that being a Grandmaster must require, demonstrates remarkable humility in describing his mental, emotional and strategic errors. Readers who can’t tell a pawn from a rook will gain much from Kasparov’s sensitive, knowing conjecture. Those who love chess or computers will enjoy a transcendent reading/thinking experience. Kasparov believes his matches with Deep Blue presaged better human/machine collaborations, which have existed since a pre-human first grabbed a rock.
Senior visiting fellow at the Oxford Martin School Garry Kasparov also wrote How Life Imitates Chess; and Winter Is Coming with Mig Greengard.