New deep learning tools, called SC2LE, hope to benchmark AI performance against human counterparts
London based technology company DeepMind has announced a partnership with Blizzard Entertainment to conduct AI experiments using StarCraft II. The company, which is part of the Alphabet group, has released a set of tools SC2LE in order to help develop artificial intelligence in areas where is it not normally designed to operate.
The company posted the information on a blog on its website, which was reported earlier today by IGN. The company believes that it is 'crucial to benchmark agent performance' in videogames that they have not specifically been designed to play and to also play against humans who are good at the game.
A machine learning API has been developed by StarCraft II developers Blizzard, which is included in the tool set, along with an open source version of DeepMind's own PySC2 with an aim to publish a joint paper with the results from the tests.
"StarCraft and StarCraft II are among the biggest and most successful games of all time, with players competing in tournaments for more than 20 years," the company wrote in its blog post. "The original game is also already used by AI and ML researchers, who compete annually in the AIIDE bot competition. Part of StarCraft’s longevity is down to the rich, multi-layered gameplay, which also makes it an ideal environment for AI research.
"The release [of PySC2] also contains a series of ‘mini-games’ - an established technique for breaking down the game into manageable chunks that can be used to test agents on specific tasks, such as moving the camera, collecting mineral shards or selecting units. We hope that researchers can test their techniques on these as well as propose new mini-games for other researchers to compete and evaluate on.
"Our initial investigations show that our agents perform well on these mini-games. But when it comes to the full game, even strong baseline agents, such as A3C, cannot win a single game against even the easiest built-in AI. One technique that we know allows our agents to learn stronger policies is imitation learning.
"Our hope is that the release of these new tools will build on the work that the AI community has already done in StarCraft, encouraging more DeepRL research and making it easier for researchers to focus on the frontiers of our field."
Deep learning and video games are not a new partnership. Nvidia also has a branch of its company dedicated to AI improvement. EA launched Seed, its own deep learning company, at E3 this year. DeepMind has previously been used with Atari's classic game Breakout to experiment with AI and reinforcement learning.