Super Mario Artificial Intelligence Learns How to Feel, Plays Own Game [Video]

A team of cognitive scientists has reprogrammed Super Mario Advance, a classic game for the Gameboy Advance portable gaming device, to give the game's main character cutting-edge artificial intelligence.

Researchers from Germany's University of Tubingen have given artificial intelligence to Super Mario, allowing the world's most iconic plumber to play his own game. The process gave Mario the ability to speak, listen, learn and even feel "emotions."

Mario is now able to respond to questions and vocal commands. In addition, the researchers gave Mario the ability to learn based on previous experiences, allowing the character to appropriately respond when given specific conditions. Mario also now has feelings, which motivate him to explore the world where he lives in.

For example, Mario learns that when he jumps and touches a coin, the coin disappears and the number of coins he has increases. In addition, Mario no longer feels hungry. This allows Mario to figure out that whenever he feels hungry, he should look for a coin and jump on it.

"We took a Mario clone that has been developed by others and then we gave the agent basic knowledge of what his behavior does," said University of Tubingen cognitive modeling head Martin Butz, one of the project's pioneers.

Butz added that the team of researchers has given Mario an internal motivated state, such as collecting enough coins while he interacts with the video game world where he is on. The researchers then give him internal needs, which are called a constant homeostatic state, such as hunger.

Whenever the equilibrium goes off balance, Mario then learns the appropriate response based on the character's past interactions with the objects in his world.

Mario has also learned to formulate his experiences into speech, with the character speaking through speech generation. Mario is also able to understand sentences, which he then includes in his knowledge base.

The knowledge base of Mario was initially empty. However, as the character interacted with the objects in his environment, Mario built up knowledge rules that can grow as big as what can be found in the real world.

Butz said that the project was partly initiated for fun, but it was also driven for the illustration of how the learning principles found in psychology and cognitive science can be used and implemented within an artificial setting.

While the Mario with artificial intelligence is only currently a prototype, Butz said that the technology can be included into any video game that features a simulated environment where a character interacts with objects.

ⓒ 2024 TECHTIMES.com All rights reserved. Do not reproduce without permission.
Join the Discussion
Real Time Analytics