The evolutionary algorithm learned to play Super Mario Kart on the SNES. (Image Credit: LF_MrL314/X)
LF_MrL314 developed an evolutionary algorithm called LeUAgi that learned how to play Super Mario Kart on SNES. The kicker here is that it doesn't require much information to play and improve. It works by learning from mistakes it may encounter in-game. For instance, the algorithm attempts to stay on the race-course track before turning corners, optimizing its playability.
"Meet LuEAgi, a program that learns to play Super Mario Kart with minimal information via an evolutionary algorithm! The best part? You can watch it learn LIVE! Come check it out; I'll be streaming for a few days, so come hang out and watch," LF_MrL314 wrote in his X post on August 31st.
MrL314 already streamed his algorithm on Twitch, and viewers can watch replays of it attempting to complete a race without driving off course. If the PC unsuccessfully turned a corner, it would respawn at the start and try again. During its learning process, the kart traveled farther and turned corners. Afterward, that data was fed into the PC, which was analyzed before adapting it for the next try.
AI in gaming has advanced over the past years, starting from basic pattern recognition to reinforcement learning and neural networks. LuEAgi's ability to function on the SNES demonstrates how AI adapts and perfects gaming systems.
Have a story tip? Message me at: http://twitter.com/Cabe_Atwell
Top Comments