Agent evaluation often focuses on the in-practice game outcomes yet typically fails to evaluate robustness to worst-case outcomes.
Presenting ISMCTS-BR, a new algorithm able to learn a best response & approximate the worst-case performance: https://t.co/AFN5XqefiY #IJCAI2022 1/
Introducing Player of Games - general and sound search algorithm that combines game theoretic reasoning and reinforcement learning. PoG learns to play both perfect and imperfect information games while bringing new SOTA result in the game of Scotland Yard https://t.co/rK74uZe2uD
@eskoil@doniveson So true. Considering how many civic employees are public facing, like police and fire, renegotiations should aim to compensate them for the even greater risks they now face. Especially considering many already faced budget cuts at the end of 2019 due to the UCP budget.
In a collaboration with Google Brain, we are releasing the Hanabi Learning Environment: a testbed for collaborative multi-agent learning and theory of mind in the card game Hanabi.
Paper: https://t.co/MTT5PvMHoP
GitHub: https://t.co/hDzYElHfQa
“It was that mystery, right from the beginning — the impossibility, yet inevitability, of an electronic brain.” - Amii’s Rich Sutton
Thrilled to see such an in-depth look at the #yeg#AI scene in this month’s issue of @AvenueEdmonton!
The global expansion of @DeepMindAI to our city is incredible news! It will enhance #YEG's reputation as an #ArtificialIntelligence hub. https://t.co/PgvDMzCZLJ
Congratulations @UAlberta! This investment will make Edmonton a leader in innovation, jobs & growth. Thanks for choosing Canada @DeepMindAI. https://t.co/LTUPJIDmjP
Tomorrow, DeepStack plays Mike McDonald for Poker Night, er afternoon, in Canada! I'm pretending the Oilers fans outside cheer for DeepStack https://t.co/dIn3GMOmId