- This AI found a flaw in a game called Q*Bert and racked up almost infinite points
- AI designed by researchers at the University of Freiburg
- Read on to access the official research paper and video
After DeepMind’s AlphaGo, the competition to make AI technology in order to play and beat games has been intense. The AI developers are given an existing and preset set of conditions and are asked to create the technology to beat it. It’s a match made in AI heaven.
So it comes as no surprise that a group of three researchers from the University of Freiburg designed one such AI. The system was being designed and trained to play games made by Atari. It came across a bug in the Q*bert game which allowed the system to rack up almost an infinite number of points.
In this game, the character jumps from cube to cube. With each jump, the color of the platforms changes and points are rewarded when all the colors are changed. But the AI found a curious flaw. According to the researchers:
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).
You can read the research paper in full here and watch the video of the AI wreaking havoc in the game below:
Our take on this
This game has been around for years but the very fact that a human never spotted this flaw and an AI did? It bodes well for evolutionary algorithms in the future.
An evolutionary algorithm puts algorithms against one another to see which one completes a task in the best possible way. It then adds a few changes to the remaining algorithms to make them better step by step.
Researchers love using machine learning on games to test how far they can take their algorithms. This trend will continue for a long time to come and we can’t wait to see how far it goes!
Subscribe to AVBytes here to get regular data science, machine learning and AI updates in your inbox!
You can also read this article on our Mobile APP