Mathematics Postgraduate Seminar
6th March 2020, 4:30 pm – 5:30 pm
Fry Building, 2.04
"We [propose] the game of Hanabi as a new challenge domain with novel problems that arise from its combination of purely cooperative gameplay and imperfect information in a two to five player setting... We believe developing novel techniques capable of imbuing artificial agents with such theory of mind will not only be crucial for their success in Hanabi, but also in broader collaborative efforts, and especially those with human partners." - Deepmind, The Hanabi Challenge: A New Frontier for AI Research (2019)
In this talk I will survey a range of strategies developed for beating Hanabi. There are many small strategies that are often employed to encode as much information as possible in a single clue such as 'finesse'. However, some of the current best strategies do not rely on these and arise by considering different problems or games based on inductive reasoning, for instance: hat puzzles. I will explain how solutions to these problems yield strategies for Hanabi and what are the mathematical and AI challenges currently being tackled by the likes of Deepmind.
I will not assume any familiarity with the game. For the curious, the rules can be found here: