# AAAI Experiences ## Submissions/Acceptance rates * 1591 submissions / 7737 reviewed submissions * acceptance rate of 20.6% * up from 7095 reviewed submissions last year (16.2% acceptance rate) ## Reinforcement Learning in Games Workshop * http://aaai-rlg.mlanctot.info/ * link with papers ![](https://i.imgur.com/cvHrqCh.png) * Search in Cooperative Partially Observable Games * FB AI Research, Hanabi * Invited talk by Amy Greenwald * Historical stuff * Atari-HEAD * New data with eye tracking on Atari * Ran some simple experiments e.g. imitation learning to test the value of the data set * Invited talk by Gerry Tesauro * Historical stuff, TD-gammon * Invited talk by Michael Bowling * Hanabi overview * Invited talk by David Silver * Got to vote between MuZero and AlphaStar * Presented MuZero * Posters * Accelerating Self-Play Learning in Go * KataGo author, lots of smaller improvements * divided poster into Go specific and general techniques * Claims 50x speedup in AlphaZero training * Analysis of Statistical Forward Planning Methods in Pommerman * hand-crafted model for model-based RL on Pommerman * Self-play Learning Without a Reward Metric * AlphaZero modification which uses ordering of game outcomes to train agents ## Reproducibility in AI Workshop * https://folk.idi.ntnu.no/odderik/RAI-2020/ * More info * Workshop chair came up with a list of guidelines for authors to follow * Also presented ML reproducibility checklist * Discussion on how to incorporate these guidelines into the submission process for future conferences * voluntary requirements show good results: * NeurIPS 2018 adoption rate <50% * NeurIPS 2019 75 ## Invited Talks * https://aaai.org/Conferences/AAAI-20/livestreamed-talks/ * all listed talks and more accessible with this link * Turing award winners * Hinton * Capsule networks * LeCun * Overview of ML topics * Promotion of self-supervised learning * Bengio * AI should take inspiration from "Thinking fast and slow" * Kautz: Third AI summer * History * Presented a few potential ethical problems * Mental privacy * Social networks * autonomous resource extraction * Debate: Academic AI researchers should focus their attention on research problems that are not of immediate interest to industry * AI History Panel: Advancing AI by Playing Games * Murray Campbell * Garry Kasparov * Michael Bowling * David Silver * Hiroaki Kitano