
Levente Kocsis, Senior Research Fellow at HUN-REN SZTAKI, was a guest on Tilos Radio’s program Természetes intelligencia: micsoda világ!, where he discussed his 2006 publication that propelled the development of AI algorithms used in Go games.
Together with his colleague Csaba Szepesvári, Kocsis made a major contribution to the explosive progress of Go-playing algorithms (including AlphaGo). In 2006, they developed the Monte Carlo Tree Search algorithm (UCT – Upper Confidence bounds applied to Trees), which was the first to start defeating human players. Incidentally, AlphaGo defeated the European champion exactly ten years ago.
Since then, Monte Carlo search has been widely applied across many domains—from language model planning tasks to bioinformatics and epidemic spread simulations. The full interview can be listened to here.