Details
The collaboration began in 2016 and produced PySC2, an open-source learning environment released in 2017, along with a large dataset of anonymized professional game replays. AlphaStar was trained first through supervised learning on human replays, then refined through reinforcement learning — a process where the AI plays repeated matches against itself and adjusts its strategy based on wins and losses. AlphaStar was tested anonymously on the public Battle.net ladder, where Blizzard gave players the option to opt out of matches against it. The research was published in the scientific journal Nature in 2019. This was a research project rather than a commercial product, but it established StarCraft II as a leading benchmark environment for AI research into complex real-time decision-making.
Have evidence about Activision Blizzard's AI practices? Submit a report.
Report a Sighting →