Slashdot Mirror
Blizzard and DeepMind Turn StarCraft II Into An AI Research Lab (techcrunch.com)
Last year, Google's AI subsidiary DeepMind said it was going to work with Starcraft creator Blizzard to turn the strategy game into a proper research environment for AI engineers. Today, they're opening the doors to that environment, with new tools including a machine learning API, a large game replay dataset, an open source DeepMind toolset and more. TechCrunch reports: The new release of the StarCraft II API on the Blizzard side includes a Linux package made to be able to run in the cloud, as well as support for Windows and Mac. It also has support for offline AI vs. AI matches, and those anonymized game replays from actual human players for training up agents, which is starting out at 65,000 complete matches, and will grow to over 500,000 over the course of the next few weeks. StarCraft II is such a useful environment for AI research basically because of how complex and varied the games can be, with multiple open routes to victory for each individual match. Players also have to do many different things simultaneously, including managing and generating resources, as well as commanding military units and deploying defensive structures. Plus, not all information about the game board is available at once, meaning players have to make assumptions and predictions about what the opposition is up to.
It's such a big task, in fact, that DeepMind and Blizzard are including "mini-games" in the release, which break down different subtasks into "manageable chunks," including teaching agents to master tasks like building specific units, gathering resources, or moving around the map. The hope is that compartmentalizing these areas of play will allow testing and comparison of techniques from different researchers on each, along with refinement, before their eventual combination in complex agents that attempt to master the whole game.
It's such a big task, in fact, that DeepMind and Blizzard are including "mini-games" in the release, which break down different subtasks into "manageable chunks," including teaching agents to master tasks like building specific units, gathering resources, or moving around the map. The hope is that compartmentalizing these areas of play will allow testing and comparison of techniques from different researchers on each, along with refinement, before their eventual combination in complex agents that attempt to master the whole game.
Never played any starcraft or WofW.
Assuming you could create some AIs that are really, really good at an RTS like SC2, I wonder if you could get them to work backwards to create (or just adjust existing) games that are more balanced as one of the major complaints about any RTS game is its balance.
... from controlling units in a RTS to controlling drones in the field of war.
Personally I welcome our robotic overlords
I'm not sure if this is good or bad for the game itself. Currently Blizzard uses cheats for "AI" opponents and in the original Starcraft game, many bugs and quirks that rudimentary classifiers found useful were then replicated by humans to make the game rather linear and boring.
Custom electronics and digital signage for your business: www.evcircuits.com
It's funny how I never stopped playing games, but I stopped being a gamer the very moment I became old. Now all I get are incredulous reactions when I talk about gaming, because nobody over the age of 25 can be a gamer.
Things fall apart; the centre cannot hold;
Mere anarchy is loosed upon the world,
The blood-dimmed tide is loosed.....
Surely some revelation is at hand;
Surely the Second Coming is at hand.
And what rough beast, its hour come round at last,
Slouches towards Bethlehem to be born?
W.B. Yeats, The Second Coming
Not slouching towards Bethlehem but lurching ever closer to Judgment Day
Pain is merely failure leaving the body
Have we found, at long last, a possible way to beat the Koreans at StarCraft?
If so, will this Make America Great Again?
can it play global thermonuclear war?
US first strike = SK mass damage and china unknown.
NK first strike = SK mass damage / Japan damage / 50-50 shot that 1 missile makes it all the way to the usa. China moves in to crush NK.
If there is an Linux API does it mean we will get the full game on Linux as well ? :
If you haven't read Charles Stross' Lobsters, take some time this weekend to do yourself the favor.
All of Accelerando is available online under the Creative Commons Attribution-NonCommercial-NoDerivs 2.5 License, but Lobsters can easily stand alone.
So, we're going to develop and train AIs using a game that focuses on developing and processing resources, building weapons and defensive systems, and coordinating troops in an assault.
Yeah. That'll be fine.