The 2012 International Go Symposium in Black Mountain, North Carolina attracted leading scholars and researchers from around the world for two days of presentations and discussions on the many aspects of the game of go. Hours of footage have now been edited down and posted online to accompany the conference papers. This 3-part series covers highlights of Symposium presentations by teachers, scientists, historians and anthropologists.
In addition to the teachers and social scientists whose presentations have been described previously in this series, several computer scientists and mathematicians brought their colleagues up to date on recent advances in those areas at the 2012 Symposium. Programmers have been trying to create strong go programs since the creation of the first computers in the 1950’s, but their efforts were so ineffective that go was dubbed “the fruit fly of artificial intelligence.” Top-level play is still an exclusive human domain, but these days computers are closing in, thanks largely to the so-called “Monte Carlo Tree Search” (MCTS) technique, which selects each move by playing out hundreds of thousands of games and choosing the move that seems to win the most games. ZenGo, an MCTS-based program, recently defeated the legendary Takemiya — twice! And in October, playing as “Zen19”, ZenGo finished nearly 200 games over a four-day period on KGS, ending with a ranking of 5D. Jacques Basaldúa, the author of GoKnot, reviews the basics of the MCTS algorithm and explains how a new filtering mechanism known as “CLOP optimization” is making programs stronger than ever (right). Francois van Niekerk also looks at MCTS, focusing on its unique ability to make use of parallelization and supercomputers.
In the world of mathematics, Prof. Elwyn Berlekamp and others have been exploring how to calculate the “temperature” of endgame moves, for instance in Berlekamp’s Chilling Gets The Last Point. Bill Spight, who wrote a chapter about Berlekamp’s “Coupon Go” in More Games of No Chance, presents some further thoughts on Berlekamp’s concept of “thermographic” analysis, and discovers surprisingly deep questions of life and death on the 3×3 board. Kyle Blocher uses combinatorial game theory in a different way, to develop a method for assessing the value of moves that he calls “miai counting”. Another aspect of the game that seems to yield to mathematical analysis is seki, as we learn from Thomas Wolf, the author of Mastering Ladders. Wolf explores how to recognize a seki when it appears, differentiating among “basic”, “linear” and “circular” types and developing a way to express seki in mathematical terms (left).
The AGA and the 2012 US Go Congress are extremely grateful to The International Go Federation for financial support that made this event possible, and to The American Go Foundation for supporting the video recording. NOTE: Links to all the videos and to associated papers, links and contact information be found at the Symposium website.