Artificial Intelligence (AI) systems are growing every single day with their applications based on deep learning. These have resulted in breakthrough approaches in the areas of understanding brain function and even autism therapy. Many efforts are being put to producing literature which can tackle some of the general public’s concerns and lack of information about what an AI can do.
A group of researchers from Georgia Tech has designed an AI system which can recreate classic 2D video games by studying the way they are played. The researchers have pulled off their project with Super Mario Bros. and Mega Man. To achieve the results the team used an engine cloning method which involved only a praser and machine’s effort to study and memorize the pixels on the screen. It also gets a lot of information about a set of concepts which explain the relationship between certain aspects of the game and a group of images compiled from the game comprising of heroes to the villains.
The AI system was able to gather information within two minutes, and even though the new version developed by the AI was not a perfect match, it was an impressive copy. Matthew Guzdial, lead researcher and Ph.D. student in computer science said in an interview, “For each frame of the video, we have a parser which goes through and collects the facts. What animation state Mario is in, for example, or what velocities things are moving at. So imagine the case where Mario is just above a Goomba in one frame, and then the next frame the Goomba Is gone. From that it comes up with the rule that when Mario is just above the Goomba and his velocity is negative, the Goomba disappears.”
The most exciting and rewarding thing about this research is that it was achieved using a method which was not depending on the code. Guzdial said, “Our AI creates the predictive model without ever accessing the game’s code, and makes significantly more accurate future event predictions than those of convolutional neural networks. A single video won’t produce a perfect clone of the game engine, but by training the AI on just a few additional videos, you get something that’s pretty close.” The machine learning system will change the concept of what is the potential for AI in the area of gaming.
The team acknowledges that more work will be needed to get done to enhance the capabilities of the system to be used with 3D platformers. Mark Riedl, associate professor of Interactive Computing and co-investigator on the project said, “The technique relies on a relatively simple search algorithm that searches through possible sets of rules that can best predict a set of frame transitions. To our knowledge, this represents the first AI technique to learn a game engine and simulate a game world with gameplay footage.”