Computer Scientists Generate A Self-Aware Mario That Can Learn And Feel

Screenshot from YouTube / Mario Lives!

For some of the world’s brightest brains, artificial intelligence (AI) could spell doom and gloom for mankind. So it’s nice to know that some researchers are managing to advance this exciting field of study whilst keeping it light-hearted. A trio of computer scientists from the University of Tübingen, Germany, has used an adaptive learning AI approach to bring one of our childhood heroes, Mario, to life. Well, kind of.

A “living and conversing” version of the agent was recently generated as part of the “Mario Lives!” project, which was created for a video competition organized by the Association for the Advancement of Artificial Intelligence (AAAI). According to AAAI, which is a nonprofit scientific society, the aim of the competition is to showcase how much fun AI is by documenting exciting advances in the field.

The idea behind this project was to develop an artificial agent who gets to know and thus becomes alive in his own world. “As most of you know, this is Mario,” a researcher explains in the video below. “But what you do not know is that this Mario has become aware of himself and his environment, at least to a certain extent.” The Mario they created can understand a large variety of sentences, answer questions, learn and even make decisions based on how he is feeling.

As explained in the video, Mario starts out with knowledge about his body and then adds additional context to his knowledge base as he learns. He also maintains emotive states which can be influenced both by the environment and by voice commands given by the user. For example, when the character was told not to be so happy, he responded in a computer-generated voice “Somehow, I feel less happy.”

Not only can Mario understand a large variety of sentences, but he can also be taught. For example, he can be told “If you jump on Goomba, Goomba dies,” but he can also learn this from experience. In the video, Mario is asked to jump on Goomba, one of his enemies, and then report back on what he learned. After landing on Goomba and realizing that he dies, Mario translates this new information into human speech and says: “If I jump on Goomba, then he certainly dies.”

Furthermore, Mario is also programmed to alter his behavior based on his mood and motivations. For example, when he is “hungry,” he searches for coins, but when he is “curious” he will spend time exploring his environment.

As Fabian Schrodt, one of the brains behind Mario Lives!, explains to The Verge, it’s not the artificial intelligence that makes the project interesting, since Mario isn’t that different to a generic computer opponent. Rather, it’s the combination of AI programming with principles from psychology that makes it unique.

Schrodt says that they are now extending this work with a follow-up project in which Mario and his companion, Luigi, are able to communicate with one another, sharing information as they go and thus teaching each other new things.

 

 

[Via The Verge, Extreme Tech]

If you liked this story, you'll love these

This website uses cookies

This website uses cookies to improve user experience. By continuing to use our website you consent to all cookies in accordance with our cookie policy.