jgravelle / so_long_sucker

0 stars 0 forks source link

Top level trolling? :P #1

Open tristan-mcinnis opened 1 week ago

jgravelle commented 1 week ago

No, an actual project based upon this...

arXiv PDF("Reinforcing Competitive Multi-Agents for Playing So Long Sucker") Learning to Play a Complex Game with Artificial Agents

Researchers studied how artificial agents can learn to play a strategy game called So Long Sucker, which involves building coalitions and betraying others. The goal was to teach the agents the rules of the game using traditional machine learning methods. To do this, they created a new version of the game with a user-friendly interface and tools for testing different algorithms. After training the agents for thousands of games, they found that while they could learn some basic strategies, they still made mistakes and required a lot of time to reach their full potential. This suggests that traditional machine learning methods may not be effective in complex social games like So Long Sucker, and that new approaches are needed to improve how quickly and accurately agents can learn.