How GTA, PacMan & other video games can help deep dive into AI?

Video games have always been an interest for AI researchers. Might be sounding unreal, but video games like Grand Theft Auto and Minecraft can actually help you a lot in learning how Artificial Intelligence works.

But why Video games?

Understanding and learning how to navigate in a 3D structure is a great tool to learn AI. As you get to learn how AI might work in complex real-life situations. In previous attempts to learn embodied cognition, there involved a lot of robots with sensors, teaching them how to operate into the real world and bumping into the things. And hence, there was a lot of maintenance work involved into it. This didn’t turn to be a huge success as there were many problems in scaling up this business. With virtual robot in a virtual world, there is obviously no maintenance required, no specifications and maintaining separate parts involved. The environment can be easily altered and manipulated. A computer can run thousands of such simulations at a time, allowing legions of virtual robots to try tasks, again and again, learning with each attempt. That kind of large-scale testing, which permits the learning process itself to be monitored and understood, is simply not practical using real machines. Hence, is a much logical approach towards learning embodied cognitive.

According to a team of from Intel and Darmstadt University in Germany, Video games are not only in the high fidelity of material appearance and light transport simulation. It is also in the content of the game worlds: the layout of objects and environments, the realistic textures, the motion of vehicles and autonomous characters, the presence of small objects that add detail, and the interaction between the player and the environment.

Back in 2015, Google AI was successfully able to win 50 Atari games.But, almost forgot it’s strategy right after the game, after it had moved on to the next game. It is not only teaching them to train it’s DeepMind artificial intelligence to remember the strategies using an algorithm called Elastic Weight Consolidation. This algorithm will help the AI to learn things instead just dumping the data. So far, it has been great at beating Breakout.

This year has been tremendous in the field of Artificial Intelligence winning games. Here is a brief throwback why:

Recently, Google’s AlphaGo software has defeated its human opponent in the first of a three-game Go match held this week at the Future of Go festival in China. While AlphaGo is making its mastery of the game look effortless, Go has typically been harder for AI to beat than chess.

Watch Udacity experts talking about this major AI win:

Microsoft’s AI is the first to reach a perfect score of 999,990 in PacMan, surpassing the best human score by 4x. The scientists used a method called Hybrid Reward Architecture, used more than 150 agents, each of which worked in parallel with the other agents to master Pac-Man.

Here’s how they did it

A team of AI researchers from Princeton University has been teaching AI how to drive with ‘Grand Theft Auto V.’ Using the graphic library in the game and injecting some additional software, providing labels that you come across while driving including pedestrians, traffic signals, cyclists, electric poles, trees and the sky. Here’s how Youtuber Python Plays explains it.

It’s quite apparent that AI is getting closer to start learning and behaving like humans. And, the future is quite near than you think. Start learning the basics of Artificial Intelligence today with Artificial Intelligence Nanodegree: