A former chess world champion has teamed up with DeepMind to find new ways to play chess

Kramnik, who retired from chess last year, agrees that his beloved sport has lost its creativity. He blames this in part on computers, whose relentless calculations create plenty of openings and defenses that big-time players are familiar with. "In quite a few games at the highest level, half of the game - sometimes the whole game - is played in memory," Kramnik said. "You don't even have to prepare yourself."

Recently, Kramnik has come up with some ideas on how to restore the human art of chess with the help of the most powerful chess computer in the world. He teamed up with DeepMind, Alphabet's artificial intelligence lab, whose researchers challenged their superhuman chess software, AlphaZero, to learn nine chess variants that allow players to generate creative new patterns.

In 2017, AlphaZero showed that it could teach itself to completely beat a chess, Go, or Japanese Go master. Kramnik said the latest findings suggest that chess has tantalizing new prospects to explore if people are willing to make a few small changes to existing rules.

The project also demonstrates a collaborative model between chess players and machines. "Chess engines were originally used to beat humans," said Nenad Tomasev, a DeepMind researcher who worked on the project. "Now, we're seeing systems like AlphaZero being used for creative exploration in collaboration with humans, rather than against them."

People have been playing chess for 1,500 years, and tweaking the rules is nothing new. No one complains that the computer makes the game boring.