Chuck Bednar for redOrbit.com – Your Universe Online
Everyone’s favorite princess-rescuing video game plumber is finally breaking free of controller-based commands and doing his own thing, thanks to a new artificial intelligence (AI) project at the Cognitive Modeling Group at Germany’s University of Tubingen.
The Mario A.I. Project essentially brings the software icon to life within the confines of his own virtual world, allowing him to think for himself. Furthermore, the project allows the character to respond to spoken instructions, and has made him aware of himself and his environment.
According to BBC News, Mario can be programmed to make decisions on what to do based on what he has been taught. For instance, the character can learn that he can definitely kill Goombas (one of the recurring in-game enemies) by jumping on it.
Furthermore, Mario will respond to how he is “feeling” based on “internal emotive states” that he has been pre-programmed with. If he is hungry, he will search for and collect coins, and if he is curious, he will explore the various levels of the Mushroom Kingdom.
“Perhaps most usefully from a gaming point of view, Mario can calculate how many moves he needs to make to reach a certain position,” the British news organization added. The students are using the project as part of an ongoing study to determine how the human mind works.
The software used for the Mario A.I. Project is Super Mario Advance for the Nintendo Game Boy Advance, and the programmers also used a speech recognition toolkit developed at Carnegie Mellon University so that the character can understand voice commands, CNET explained.
When phrases from the toolkit’s language tree are spoken to Mario, he chooses from a range of different possible actions based on what he has learned. He can also plan his actions several steps in advance, calculating the number of jumps, the jump height, the direction and distance he needs to travel in order to reach a particularly difficult destination, the website added.
“The concept of a computer-driven Mario isn’t completely new, and it’s doubtful that this AI will get so smart that it’s performing speed runs,” said Engadget.
However, the website added, the project demonstrates “how cognitive computing can help in games and other situations where you’d want software to adapt to unexpected conditions. If techniques like this catch on, you could see in-game enemies that learn to counter your moves, or autonomous cars and robots that can adapt to new hazards based on your suggestions.”
A video created by Stephan Ehrenfeld, Fabian Schrodt, and Professor Dr. Martin V. Butz at the university, “Mario Lives! An Adaptive Learning AI Approach for Generating a Living and Conversing Mario Agent,” was uploaded to YouTube last week. That short film has been entered in the ninth annual People’s Choice Award for the AI Video Competition.
Voting for the AAAI Video Competition on January 27 at midnight (CST), and the presentation ceremony will be held in conjunction with the AAAI-15 conference in Austin, Texas on January 29. In addition to the People’s Choice Award, a panel of AI experts will honor winners for Best Video and Best Student Video, according to the organization.
—–
Follow redOrbit on Twitter, Facebook, Instagram and Pinterest.
Comments