麻豆传媒

A new version of the ancient Chinese board game Go that uses quantum entanglement to add an element of randomness could make it a tougher test for artificial intelligences than regular board games.

鈥淏oard games have long been good test beds for AI because these games provide closed worlds with specific and simple rules,鈥� says Xian-Min Jin at Shanghai Jiao Tong University in China. In Go, players take turns to place a stone on a board, trying to surround and capture the opponent鈥檚 stones. There are 10¹⁷¹ possible states of聽the board, compared with around 10⁵⁰ for chess, making Go a far more complex game.

AlphaZero, an AI created by聽DeepMind, has mastered this聽complexity to become the聽world鈥檚 best Go player, but now Jin and his colleagues have developed a new version of Go that is even more complex.

Quantum Go can be played on an ordinary board, but also requires a computer to record the state of the game and equipment to generate pairs of聽quantum-entangled photons. Each player places two stones at once on their turn, representing a superposition of two possible locations of a single quantum stone. When a new stone is put next to either of those locations, the quantum state of a pair of entangled photons is measured to determine the original stone鈥檚 location, collapsing the聽superposition. And then the聽other stone is removed.

This added randomness makes the game more complex. Players can also choose to adjust the probability of their quantum stone appearing in one location, meaning it doesn鈥檛 have to be 50-50. This allows one player to聽have more information than their opponent about where a聽stone is likely to end up.

AIs can beat humans at other such 鈥渉idden information鈥� games, like certain variants of poker, but the complexity of Go brings a new challenge.

鈥淨uantum Go adds a lot more聽complexity to regular Go聽by expanding the possible ways the board can change on a player鈥檚 turn and that adds a lot more mental load to planning,鈥� says Mike Cook at Queen Mary University of London. 鈥淵ou end up planning for many different possible futures鈥�, which takes more computing power.

鈥淎s artificial intelligences surpass humans in various fields, the quantum regime may be the only space where human consciousness and intuition may beat the huge computing power of artificial intelligence,鈥� says Jin. The team has yet to train an AI to play quantum Go,聽so we don鈥檛 know if it would聽beat human players.

Cook says harder board games may not be the way forward. 鈥淎I research isn鈥檛 just about finding the most complex, tangled problem possible, it鈥檚 also about finding problems that are deceptively simple and figuring out why AIs are so bad at them,鈥� he says. 鈥淨uantum Go sounds like a very fun game, but I don鈥檛 think we鈥檇 learn a lot about AI by beating it聽鈥� although I鈥檇 still very much enjoy seeing it happen.鈥�

Reference: