Watch AI Defeat this Classic Video Game by Finding a Glitch
German researchers developed an AI that unexpectedly won against a classic video game by finding a loophole.
Within the last two years, the public has seen artificial intelligence systems defeat some of the toughest video games and board games around. They've even taken out nearly every human player who dared go up against them. But this latest win for AI has researchers discussing whether or not the AI learned to 'cheat' by breaking the game.
Three researchers from the University of Freiburg in Germany developed an AI that could navigate through popular video games. More specifically, the team wanted to see how AI handled classic 8-bit video games like those in the Atari era. So they let their AI take on popular retro game Q*bert.
Normally in the game, players jump from platform to platform to earn points and make it to the next level. It's an addictive game that cemented itself rather quickly alongside fan favorites like Galaga and other classic arcade games.
However, despite decades of players, no human player ever stumbled upon one particular glitch.
In the paper, the researchers explained the AI's new way to play:
"First, it completes the first level and then starts to jump from platform to platform in what seems to be a random manner. For a reason unknown to us, the game does not advance to the second round but the platforms start to blink, and the agent quickly gains a huge amount of points (close to 1 million for our episode time limit)."
The AI program is part of larger research about "evolutionary algorithms." This is a system in which algorithms compete head-to-head to see who wins out -- a type of AI natural selection. This way, the algorithms continue to get better. With that in mind, the researchers pointed out that the AI that defeated Q*bert isn't playing the game the same way a human player would.
Fans of Q*bert quickly took notice saying that after decades of playing the game, they never once spotted the glitch. Even Q*bert designer Warren Davis said, "This certainly doesn't look right." Davis, who didn't' work on this particular version of the game, noted that he didn't think "you'd see the same behavior in the arcade version."
Via: Patryk Chrabaszcz