Atari video game Q*bert was beaten by a program of Artificial Intelligence, which has exploited a loophole that had never been discovered.
The AI program used trial and error to discover a strangeness in the code of the game that allows you to mark a large amount of points.
No human player of Q*bert is believed to have ever discovered the tricks used to win.
The AI program was let loose on the video game by German researchers, who are developing code that can learn.The suicide strategy
Video games have proved popular with researchers because they are limited to worlds in which success (high scores) and failure (losing the game) are easy to evaluate. This can help to refine THE programs because of those who obtain the largest number of points and lose the least are likely to be better learners.
Patryk Chrabaszcz, Ilya Loshchilov, and Frank Hutter University of Freiburg to allow the most basic TO the programs loose on the classic Atari video games as part of the work on what are known as “evolutionary algorithms”.
As the name suggests, this involves the generation of a lot of algorithms, since those work better, and then change or modify in small ways to see if you get better or worse.
These evolutionary methods are in contrast with another widely used approach known as “deep reinforcement learning” that mimic biological neural networks, and allow them to learn for themselves. The most well-known of these systems is Google’s Deep Mind.Different strategies
In Q*bert, players are presented with a pyramid made of cubes, on which you must jump. Landing on top of a cube changes its color. The player must change all the cubes of color without being caught by the game’s enemies.
Rather than the original, the researchers used an updated version of the game, and the other seven, to make it easier for them TO create to try out different strategies.
Q*bert, said the researchers, the code TO found two particularly interesting solutions”.
One revolved around a game of bug that has seen the player-controlled jump from a cube to cube in a seemingly random way. However, they found this caused the cube starts to blink and award the player with a large amount of points.
A video posted by Mr. Chrabaszcz shows the controlled by the player to get a lot of points in just 10 minutes.
Warren Davis, who worked on the original arcade version of Q*bert, said that he was not familiar with porting code, but he added, “This certainly does not look right, but I don’t think that you’d like to see the same behavior in the arcade version.”
Another novel strategy for the infinite tempting Q*bert to commit suicide. Every time this has happened, the program has received a sufficient number of points for another life, then it may repeat the cycle.
In their research paper, the team said that the success demonstrated by their “basic” algorithm, has shown the promise of this branch of AI, and could be considered as potentially competitive modern approach deep reinforcement learning algorithms”.