A ten person team building Skyrim? Google's Project Chimera set to revolutionize Stadia game development through machine learning
In a magazine interview, Google's Erin Hoffman-John, Stadia's head of creative for R&D claimed that Project Chimera, a Google machine learning solution, could be used to help small independent teams develop massive open world titles without compromising on scope and detail. Hoffman-John said this to MCVUK magazine: "What if a team of 14 people could make a game the scale of World of Warcraft?...they rely on a lot of heavy, repetitive content creation...like 70 percent content and 30 percent or less code."
Project Chimera is a GAN (Generative Adversarial Network), functionally similar to ESRGAN, a neural network modders have been using to upscale textures in older games. Project Chimera has much wider applications, though. By training it on game rules, assets, and gameplay, developers should be able to use the GAN to generate world spaces, NPCs, and quests. Hoffman-John also stated that Project Chimera could be used for playtesting. It could potentially give small development teams as much feedback on bugs and balancing issues as hundreds of human beta testers.
While Google Stadia itself is in rocky waters following a disastrous launch, Project Chimera and machine learning in general, will likely play a much greater role in videogame development. We'd better get ready for Elder Scrolls games dreamt up by AI.