Researchers create AI Mario that can think for himself… now let’s try him for his crimes!
German researchers create an AI Mario that can feel and think for itself. Hopefully, they will instill accountability to hold him responsible for his crimes.
A group of German researchers working out of the University of Tubingen have developed a way to teach Mario how to think for himself, according to a demo video captured by IGN.
The video is titled “Mario Lives! An Adaptive Learning AI Approach for Generating a Living and Conversing Mario.” That’s a mouthful, but the idea is that the research team’s Mario can answer simple questions, perform certain tasks, and learn based on what he sees.
One of the examples the video highlights shows Mario facing off against a goomba. The researchers first ask Mario what he knows about goombas, to which Mario pulls a Jon Snow, replying that he knows nothing of them. The researchers then command Mario to kill, and Mario goes off to do some killing.
It’s a bit unsettling.
Mario looks for an enemy, zeroing in on the goomba without any input from the researchers, and jumps on it, killing it dead. The researchers then ask Mario what he knows about goombas, to which he replies “If I jump on goomba, then maybe it dies.”
Mario learns from his actions. After killing more goombas, he later answers that same question by saying “If I jump on goomba, it certainly dies.”
It’s more than just storing experience though. The researchers also show that after time, Mario can make decisions based on those previous choices. To Mario, coins are food. So when he becomes “hungry,” he will seek those coins out. This makes him happy, which encourages him to seek more. To do that, he begins to explore. On the flipside, when he feels fear, he will presumably kill the things causing him to feel that fear.
Let that sink in for a second.
This concept of seeking out things that an AI deems as good and bad is one of the building blocks for Artificial Intelligence programming. It’s also part of the reason that some of the smartest people in the world, like Elon Musk and Stephen Hawking, are concerned about the rise of AI.
Perhaps instead Mario will learn compassion. Finally.
When you think about the original Super Mario Bros., Mario is something of a bastard. Sure, Princess Peach is kidnapped and he is trying to find her, but in that first game the only reason we know the goombas are enemies is because they can hurt Mario. That’s it.
The goombas could just as easily be the wildlife of the Mushroom Kingdom. They tend to keep to themselves, harmlessly walking back and forth until Mario decides that they are in his way. Instead of just jumping over them, he decides to crush them. Some, he even kills so he can steal their shell to use against other animals. Technically, that makes Mario a murderer and a thief.
Maybe the Hammer Brothers are the real heroes of the game, as they knowingly stand against the horror of a psychopath that can grow when he takes “special” mushrooms, who can even throw fire. Maybe they are buying time for the Hammer Mother and the Hammer children to run for their lives before the onslaught of the murderous plumber.
Mario has a lot to answer for. And now maybe he will finally be held accountable for all the baby goombas left without parents.
You’re a monster, Mario. And now maybe you can learn to feel bad about it.