Super Mario World AI Listens To Your Advice And Learns How To Play

The red plumber has become self-aware. Mario AI, a project created by students from the University of Tubingen in Germany, makes Nintendo's mascot both aware of his environment and responsive to your advice.

Mario's behaviour is influenced by a few different criteria, namely the level layout he learns, what you teach him about the world (for example, he will stomp on Goombas when you have told him that it will definitely kill them), and even his own hunger and feelings. You'll see him start to explore more 'off the beaten track' routes through a level if he's feeling curious, and he'll go find some coins when feeling hungry.

An AI-controlled Mario is nothing new, and don't expect this to be a fast way to complete games in the near future. However, this kind of computation demonstrates how software can adapt to conditions if given the programmed ability to do so. it also shows how robots could respond re-actively to your advice and direction. Simply put, think of this as the beginning of iRobot.

In the spring (summer) term of 2014, nine students got together at the University of Tübingen, Germany, to enhance and further develop the Mario AI presented in this video under our supervision. Interfaced with the Java-based Mario simulation, Mario can only execute those actions that also a normal player could execute.