Blizzard Entertainment's classic strategy game StarCraft 2 is now being used as a basis for improved AI learning. The news was announced by Deepmind - a leading reseracher into artificial intelligence and its potential for positive use.
Deepmind partnered up with Blizzard Entertainment to create a set of tools called SC2LE to accelerate AI research specifically within the popular strategy game. Blizzard is also making a dataset of anonymised game replays that will increase to 'more than half a million' in the following weeks.
In Deepmind's post on its website, the company explains that part of the Star Craft 2's draw is the prevalence of 'sub-goals' within the game and multi-layered gameplay.
In the video above, you can see the difference between 'trained' and 'untrained' AI agents. While AI agents usually perform tasks well in isolated mini-games, they struggle to perform tasks when exposed to the full game. The best way to train AI is through 'imitation learning' - learning from human examples, and the huge amount of replays that Blizzard is able to share from StarCraft's massive player-base.
Read more: 12 Things You Didn’t Know About Iron Ma