CHESS HAS A reputation for cold logic, but Vladimir Kramnik loves the game for its beauty.
“It’s a kind of creation,” he says. His passion for the artistry of minds clashing over the board, trading complex but elegant provocations and counters, helped him dethrone Garry Kasparov in 2000 and spend several years as world champion.
Yet Kramnik, who retired from competitive chess last year, also believes his beloved game has grown less creative. He partly blames computers, whose soulless calculations have produced a vast library of openings and defenses that top-flight players know by rote. “For quite a number of games on the highest level, half of the game—sometimes a full game—is played out of memory,” Kramnik says. “You don’t even play your own preparation; you play your computer’s preparation.”
Wednesday, Kramnik presented some ideas for how to restore some of the human art to chess, with help from a counterintuitive source—the world’s most powerful chess computer. He teamed up with Alphabet artificial intelligence lab DeepMind, whose researchers challenged their superhuman game-playing software AlphaZero to learn nine variants of chess chosen to jolt players into creative new patterns.
In 2017, AlphaZero showed it could teach itself to roundly beat the best computer players at either chess, Go, or the Japanese game Shogi. Kramnik says its latest results reveal beguiling new vistas of chess to be explored, if people are willing to adopt some small changes to the established rules.
The project also showcased a more collaborative mode for the relationship between chess players and machines. “Chess engines were initially built to play against humans with the goal of defeating them,” says Nenad Tomašev, a DeepMind researcher who worked on the project. “Now we see a system like AlphaZero used for creative exploration in tandem with humans rather than opposed to them.”
People have played chess for around 1,500 years, and tweaks to the rules aren’t new. Nor are grumbles that computers have made the game boring.
Chess spread rapidly around 500 years ago after European players promoted a slow-moving piece into the powerful modern-day queen, giving the game more zip. In 1996, one year before IBM’s Deep Blue defeated Kasparov, chess wunderkind-turned-fugitive Bobby Fischer called a press conference in Buenos Aires and complained that chess needed a redesign to demote computer-enhanced memorization and encourage creativity. He unveiled Fischer Random Chess, which preserves the usual rules of play but randomizes the starting positions of the powerful pieces on the back rank of the board each game. Fischer Random, also known as Chess960, slowly earned a niche in the chess world and now has its own tournaments.
DeepMind and Kramnik tapped AlphaZero’s ability to learn a game from scratch to explore new variants more quickly than the decades or centuries of human play that would reveal their beauty and flaws. “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,’” says Tomašev.
AlphaZero is a more flexible and powerful successor to AlphaGo, which laid down a marker in AI history when it defeated a champion at Go in 2016. It starts learning a game equipped with only the rules, a way to keep score, and a preprogrammed urge to experiment and win. “When it starts playing it’s so bad I want to hide under my table,” says Ulrich Paquet, another DeepMind researcher on the project. “But seeing it evolve from a void of nothingness is exciting and almost pure.”
In chess, AlphaZero initially doesn’t know it can take an opponent’s pieces. Over hours of high-speed play against successively more powerful incarnations of itself, it becomes more skilled, and to some eyes more natural, than prior chess engines. In the process it rediscovers ideas seen in centuries of human chess and adds flair of its own. English grandmaster Matthew Salder described poring over AlphaZero’s games as like “discovering the secret notebooks of some great player from the past.”
The nine alternative visions of chess that AlphaZero tested included no-castling chess, which Kramnik and others had already been thinking about, and had its first dedicated tournament in January. It eliminates a move called castling that allows a player to tuck their king behind a protective screen of other pieces—powerful fortification that can also be stifling. Five of the variants altered the movement of pawns, including torpedo chess, in which pawns can move up to two squares at a time throughout the game, instead of only on their first move.
One way of reading AlphaZero’s results is in cold numbers. Draws were less common under no-castling chess than under conventional rules. And learning different rules shifted the value AlphaZero placed on different pieces: Under conventional rules it valued a queen at 9.5 pawns; under torpedo rules the queen was only worth 7.1 pawns.
DeepMind’s researchers were ultimately more interested in the analysis of the other great chess brain on the project, Kramnik. “This is not about numbers, but whether it is qualitatively, aesthetically pleasing for humans to sit down and play,” says Tomašev. A technical paper released Wednesday includes more than 70 pages of commentary by Kramnik on AlphaZero’s explorations.
Kramnik saw flashes of beauty in how AlphaZero adapted to the new rules. No-castling chess provoked rich new patterns for keeping the king safe, he says. A more extreme change, self-capture chess, in which a player can take their own pieces, proved even more alluring. The rule effectively gives a player more opportunities to sacrifice a piece to get ahead, Kramnik says, a tactic considered a hallmark of elegant play for centuries. “All in all it just makes the game more beautiful,” he says.
Kramnik hopes AlphaZero’s adventures in alien forms of chess will convince players of all levels to try them. “It is our gift to the world of chess,” he says. Now could be an opportune moment.
Chess has been gaining popularity for years but experienced a pandemic boost as many people sought new intellectual stimulation, says Jennifer Shahade, a two-time women’s US chess champion. Interest in Chess960 has grown too, suggesting an appetite for new types of play, including from some superstars. Later this week, Shahade will provide commentary for a Chess960 tournament including world number one Magnus Carlsen and Kasparov, the former champ.
Like Kramnik, Shahade saw things to like in several variants AlphaZero tested, even if changes like allowing pawns to move sideways felt “mind-bending.” If any gain traction, some players will still want to lean on computers and deep research to get ahead, but resetting the cycle could be fascinating to watch. “The discoveries would feel fresh—it could be very exciting and benefit a different type of player,” says Shahade, who is also women’s program director at the US Chess Federation.
DeepMind and Kramnik’s project might also encourage computer chess to get more creative, now that machines are unbeatable. “Instead of making computer chess stronger and trashing humans, we can focus on chess as an art in the form of a game,” says Eli David, a researcher at Bar-Ilan University in Israel who has built machine-learning-powered chess engines of his own. One grad student in his lab is working on chess software that learns to mimic the style of a particular player, which could make it possible to ask a machine what a favorite grandmaster past or present would do in a particular situation.
Kramnik’s experience suggests that having humans work with, not against, machines can expand the emotional as well as technical experience of the game. AlphaZero took him to places outside even his vast understanding. “After three moves you simply don’t know what to do,” he says. “It’s a nice feeling, like you’re a child.”