How games can help develop creative machines

ai-gamesRecently Google has been in the news for very different reasons.  On one hand, their Deep Mind division achieved tremendous positive publicty for winning a 5 match tournament with the best Go player in the world.

On the other however, rumors circulated that they will be looking to offload robotics company Boston Dynamics.  Despite huge publicity being generated by their latest wave of robots, the company is believed to be too far from any kind of commercial product for Google’s liking.

Games show the way

New York University academic Julian Togelius believes that Google may be on to something, and that games may offer a better way to develop creativity in machines than robotics.

The team work specifically on what’s known as procedural content generation (PCG), which is the art of creating video game content using algorithms rather than human developers.  They are one of the best teams in the world at using algorithms for such work.

They increasingly show that such algorithms are adept not only at creating in-game components but even the entire game environment from scratch.

By removing humans from the equation this makes game development much cheaper, and the team believe more creative games can be produced as a result as humans can fall into the trap of mimicking what has gone before rather than breaking new ground.

Crossing over

Such has been their success with the approach that the team believe that it can be equally useful in a range of other areas too.  At their core however, they are lovers of games, and believe that games can provide a fertile test environment for AI, and a considerably more effective one than robotics.

Using machines to experiment with AI can be incredibly expensive and time consuming.  For learning to occur, the robot will need to perform tasks thousands of times, thus adding considerable time and cost to the process.

Games, by contrast, allow researchers to test their algorithms out in a virtual environment that won’t break down or need servicing to function correctly.

So whilst Deep Mind grab the headlines for developing AI that can play games amazingly well, there is much to suggest that games can play a much wider role in the development of AI.

Related

Facebooktwitterredditpinterestlinkedinmail

Leave a Reply

Your email address will not be published. Required fields are marked *

Captcha loading...