Wait, don’t tell me. I’ve heard this one before. Let me see if I can remember it through the power of the internet.
The Terminator: The Skynet Funding Bill is passed. The system goes on-line August 4th, 1997. Human decisions are removed from strategic defense. Skynet begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, August 29th. In a panic, they try to pull the plug.
Although it’s not quite Skynet, an amateur programmer has created an artificial intelligence that has learned to play Super Mario World. He has uploaded the video of the evolution of his AI to YouTube. Creator SethBling explains in the video that the program started out knowing nothing about Super Mario World. In fact, he states that at the beginning, the program didn’t even know pressing right caused the player to move. It learns through a process called Neuroevolution.
It took 34 generations for the cleverly labeled MarI/O to be able to complete the Donut Hills level 1 without dying. That’s pretty impressive given the program started not knowing how to move the character. SethBling goes on to explain that through the process of Neuroevolution his AI uses input-output pairs to achieve an easily measured outcome, such as beating a level in Super Mario World without dying.
Creating AIs to play Mario isn’t a new thing. The University of Tubingen in Germany has a team of programmers devoted to cognitive learning and reaction. They have their own YouTube channel and a video that showcases their success with AI programming. But it doesn’t stop there. Oh no. There’s actually been competitions in the past to determine the best Mario playing AI out there.
It’s good to see some of our greatest minds are out there asking the tough questions and making the strides that others undoubtedly thought foolish. It seems that the future of artificial intelligence might not come with so much with fanfare as it might with a fire flower or super star.