3 min read
Video: Three unorthodox uses for video game AI
While we typically use AI to control our non-player characters, the developers in this GDC session prove that AI can be used for far more unusual and esoteric purposes.
When it comes to video games, we typically think of AI as a collection of tools that control our non-player characters. But that doesn't cover the whole picture -- developers can actually use game AI for far more esoteric purposes. During a robust, three-part panel at this year's Game Developers Conference in San Francisco, developers from Double Fine, Gas Powered Games, and Havok took a moment to showcase the unusual ways AI in which can solve tough development challenges. Double Fine's Chris Jurney opened the talk with a brief discussion on how he and his team devised a new AI solution that helped the Kinect sensor detect up to six individual people for the augmented reality toy Double Fine Happy Action Theater. Typically, the Kinect can only pick up a maximum of two player skeletons, but Double Fine's 'blob-detecting' AI technique helped the studio work around the camera's predefined limits. Next up, Mike "Sorian" Robbins of Gas Powered Games detailed the intricate neural networks that power the complex strategy game Supreme Commander 2. These complex AI systems allowed the game's numerous units to make decisions on their own, and added a new layer of complexity to the virtual battlefield. Finally, Havok software developer Ben Sunshine-Hill closed out the talk with an overview of a new approach to LOD scaling for character AI. This new solution makes resource distribution far more efficient across video game entities, and lets teams get just a bit more out of their game's performance. All of these techniques demonstrated just a few of the problems developers can solve when they approach AI in unusual ways, and you can check out this informative presentation in full by watching the above video, courtesy of the GDC Vault.