Go is an abstract strategy game from antiquity, so complex that the number of board positions far exceeds the number of atoms in the universe. After Deep Blue defeated the world chess champion Garry Kasparov in 1997, it was considered a holy grail of modern computing research to create a computer that can beat a human Go professional. Up until 2015, all the “world champion” Go programs still performed substantially below human professional level. But it all changed in March 2016 when AlphaGo scored 4-1 against a top Go professional Lee Sedol, a feat previously believed to be impossible for at least another decade.
The most recent breakthrough in Go computing research occurred when the Monte Carlo simulation was combined with deep convolutional neural network, resulting in an unprecedented leap forward in Go-playing power. This has immense implications not only in Go computing research, but for deep learning technology as a whole. Subsequently, in January 2017, AlphaGo-Master continued with a 60-0 perfect record against top human Go professionals. Few would doubt the significance of AI in the future of Go.
Next generation Go engines have the opportunity to tackle unanswered questions left by AlphaGo, as well as to explore uncharted territories. Humans have nurtured and toiled away at Go for millennia, with the advent of AI Go, immediate applications include: augmented live commentary of human Go matches, more accurate and standardized Go ranking systems, and revolutionary impact on the learning and teaching of Go, to name but a few.
2017 marks the inception of The World AI Go Open, where research teams from across the globe gather to compete in a formal tournament setting. The event is hosted by the International Go Federation, the Chinese Go Association, the Inner Mongolia Dept. of Athletics, and the Ordos City. The winner will take home the title “The 1st World AI Go Open Champion”. We hope this tournament will provide impetus and inspirations for AI not only in Go but in other areas as well.