As you heard last year, DeepMind and Blizzard has co-operated to work on AI research for the one of the most famous and successful game, Starcraft II.
After about one year, the game, Starcraft II, has been updated to open access for developers to research AI for the game. Compared to two previous games, Atari and Go, Starcraft is a much more complex game, which requires a lot of patterns, training. Therefore, by solving the AI for this game will help a lot of solving and develop more complex systems for future.
On the annoucement, the partnership releases a package for Starcraft II AI research environment, SC2LE. This package includes:
- A Machine Learning API developed by Blizzard that gives researchers and developers hooks into the game. This includes the release of tools for Linux for the first time.
- A dataset of anonymised game replays, which will increase from 65k to more than half a million in the coming weeks.
- An open source version of DeepMind’s toolset, PySC2, to allow researchers to easily use Blizzard’s feature-layer API with their agents.
- A series of simple RL mini-games to allow researchers to test the performance of agents on specific tasks.
- A joint paper that outlines the environment, and reports initial baseline results on the mini-games, supervised learning from replays, and the full 1v1 ladder game against the built-in AI.
Agent is designed to train from the game datasets and have ability to learn more tasks. It’s pretty complicated because, on a small square of an unit like 84×84, there might have more than 100 actions can be taken.
Well, with the release of Starcraft API, I believe there are more interesting things to come for AI development and progress.
Please check out detail their pages:
- DeepMind and Blizzard open StarCraft II as an AI research environment
- The StarCraft II API Has Arrived
Or just grab the SCII API and try it now: https://github.com/Blizzard/s2client-proto
