I am sharing this work I learned about a few years ago as it strongly pertains to MAS. The team at Google Deepmind worked on a series of agents to play the game Starcraft 2 and trained them to the point where they could routinely beat the best of the best in the professional scene [1]. This is fascinating as complex games such as Starcraft 2 involve much more abstract thinking and future planning, and the need for MAS in this sort of setting is crucial. Games such as these are not bound by a board and are limited by a set of "legal" moves, making this work so interesting. I have added the team's full paper here, and this reminded me of the advancements of OpenAI Five when they beat the world champions in Dota 2 [2],[3]. 

Another piece of work by Deepmind that should be brought up is AlphaGo. While the architecture for AlphaGo is not based on MARL, it did set in motion the basis for AlphaStar and other related projects at Deepmind. Additionally, many researchers from our city at the University of Alberta participated in the project. It is quite well documented in their film, which is available for free on YouTube [4]. It is a good watch without over-the-top technical terms, which would make it broadly appealing to anyone in this class.

Sources:

[1] “AlphaStar: Grandmaster level in StarCraft II using multi-agent reinforcement learning,” Google DeepMind. Accessed: Dec. 25, 2023. [Online]. Available: https://deepmind.google/discover/blog/alphastar-grandmaster-level-in-starcraft-ii-using-multi-agent-reinforcement-learning/

[2] O. Vinyals et al., “Grandmaster level in StarCraft II using multi-agent reinforcement learning,” Nature, vol. 575, no. 7782, pp. 350–354, Nov. 2019, doi: 10.1038/s41586-019-1724-z.

[3] “OpenAI Five defeats Dota 2 world champions.” Accessed: Dec. 25, 2023. [Online]. Available: https://openai.com/research/openai-five-defeats-dota-2-world-champions

[4] AlphaGo - The Movie | Full award-winning documentary. Accessed: Dec. 25, 2023. [Online Video]. Available: https://www.youtube.com/watch?v=WXuK6gekU1Y