AI researchers are beginning to turn to video games to demonstrate how their AI has progressed over time. Last year, Elon Musk-backed OpenAI successfully defeated professional players in a game of DOTA 2 at The International 8. Although Open AI played a modified version of the game with a limited hero pool, some of the behaviours that were deemed unconventional by players helped the bot snag their victory. Today, Google’s own DeepMind finished a StarCraft II Demonstration where its AlphaStar AI agent successfully defeated professional gamers 10-1.
Today’s demonstration showed off AlphaStar competing and winning against two professional players. Each played a five-game series, with DeepMind’s AI sweeping all ten rounds. DeepMind was involved with Starcraft 2 as early as 2017, shortly after beating the Go world champion. Blizzard created StarCraft II Learning Environment (SC2LE) for researchers and developers. DeepMind chose Starcraft 2 as the game is a “grand challenge” for how successful AI agents are at multiple tasks that have to be balanced concurrently in real-time. Skills needed to win include Game theory, Imperfect information, Long term planning, Real-time, and Large action space. According to DeepMind’s blog post:
Related Google’s Tacotron 2 Text To Speech AI Produces Sounds Indistinguishable From Human Speech
AlphaStar’s behaviour is generated by a deep neural network that receives input data from the raw game interface (a list of units and their properties), and outputs a sequence of instructions that constitute an action within the game. AlphaStar also uses a novel multi-agent learning algorithm. The neural network was initially trained by supervised learning from anonymised human games released by Blizzard. This allowed AlphaStar to learn, by imitation, the basic micro and macro-strategies used by players on the StarCraft ladder. This initial agent defeated the built-in “Elite” level AI – around gold level for a human player – in 95% of games.
No, we don’t have to worry about our AI overlords taking over e-sports just yet. Games such as Starcraft and DOTA help programmers model AI that can solve real-world problems based on what it ‘learned’ from the experience. The games were live streamed on YouTube and Twitch, and you can check it out in the video below.
Source: Deepmind
