Solving Imperfect-Information Games with Tuomas Sandholm - NIPS ’17 Best Paper - TWiML Talk #99

This article discusses an interview with Tuomas Sandholm, a Carnegie Mellon University professor, about his work on solving imperfect-information games. The focus is on his 2017 NIPS Best Paper, which detailed techniques for solving these complex games, particularly poker. The interview covers the distinction between perfect and imperfect information games, the use of abstractions, and the concept of safety in gameplay. The paper's algorithm was instrumental in the creation of Libratus, an AI that defeated top poker professionals. The article also includes a promotional announcement for AI summits in San Francisco.

• •Tuomas Sandholm's research focuses on solving imperfect-information games.
• •The NIPS '17 Best Paper detailed techniques used in Libratus, an AI that beat top poker pros.
• •The interview covers concepts like perfect vs. imperfect information, abstractions, and safety in gameplay.