Super-Mario-World

In the video below you can see how pretty good player passes gamer level in Super Mario World. This gamer is not a person, but software that use evolutionary principles to learn how to play. Artificial Intelligence, which have name MarI/O, has managed to find the correct principle of passing a level only after 34 attempts.

It’s an impressive amount, especially considering that the software did not know what the core mechanics of the game, in which direction should move or even looks like the end of the level. The main parameter that has helped the learning is similar to the level of health in a lot games. It is directly connected to the direction of movement, so if AI go in the wrong direction, it will report a decline in positive parameter and know that we must rectify its decision. The next steps in the game software achieves the principle of trial and error, the actions leading to positive results are saved and used again. This principle is the basis of learning system NeuroEvolution of Augmenting Topologies, which is inspired by the evolutionary process and mimics the principles of the human brain.

https://i0.wp.com/gizbrain.com/wp-content/uploads/2015/06/Super-Mario-World.jpg?fit=640%2C439https://i0.wp.com/gizbrain.com/wp-content/uploads/2015/06/Super-Mario-World.jpg?fit=190%2C150Michael K.GAMINGgamesIn the video below you can see how pretty good player passes gamer level in Super Mario World. This gamer is not a person, but software that use evolutionary principles to learn how to play. Artificial Intelligence, which have name MarI/O, has managed to find the correct principle of...