DeepMind AI AlphaStar Wins 10-1 Against 'StarCarft II' Pros (newscientist.com)

In a series of matches streamed on YouTube and Twitch, DeepMind AI AlphaStar defeated two top-ranked professionals 10-1 at real-time strategy game StarCraft II. "This is of course an exciting moment for us," said David Silver at DeepMind in a live stream watched by more than 55,000 people. "For the first time we saw an AI that was able to defeat a professional player." New Scientist reports: DeepMind created five versions of their AI, called AlphaStar, and trained them on footage of human games. The different AIs then played against each other in a league, with the leading AI accumulating the equivalent of 200 years of game experience. With this, AlphaStar beat professional players Dario Wunsch and Grzegorz Komincz -- ranked 44th and 13th in the world respectively. AlphaStar's success came with some caveats: the AI played only on a single map, and using a single kind of player (there are three in the game). The professionals also had to contend with playing different versions of AlphaStar from match to match. While the AlphaStar was playing on a single graphics processing unit, a computer chip found in many gaming computers, it was trained on 16 tensor processing units hosted in the Google cloud -- processing power beyond the realms of many.

Comments

Great news

[110010001000]

Computers are good at playing computer games with a strict ruleset. Imagine that.

Re:Great news

[Anonymous Coward]

Are you saying it doesn't?

In competitive StarCraft, you take into consider build orders where you need to have done a specific set of steps in a specific amount of time. If you fail to do it, it puts you behind the opponent and will cause you to lose the game.

It's so meticulous that seasoned casters can actually tell a build based on how many workers a player has in a minute.

Just look at the Liquipedia entry for Protoss build orders and you'll see what I mean. And that's just one race out of three.

Re:Still Cheating

[The Evil Atheist]

There's no viable route from this to general purpose AI

How do you know? Is there a roadmap that says all AI must progress in this way? If there were such a roadmap, why do we still need research then? Do you even understand why we research things we don't have adequate knowledge/experience?

It did not acquire any actual understanding of Starcraft mechanics

If it beat professional human players, then yes it did acquire an actual understanding of Starcraft mechanics. In fact, a better understanding than humans.

If you give it Warcraft instead, it'll take another several months of work from a team of very intelligent humans to make it good at it. In fact, I'll bet a big enough balance patch will cause it to have to throw out everything it's learned.

So what? The fact it wasn't even able to do something like Starcraft before. Do you not understand that technological progress isn't linear? How long ago were people saying AI can't play Go. And how long after beating humans at Go that people like you were saying AI can't beat humans at things that require imperfect knowledge? Do you really think DeepMind started from scratch after conquering Go, or maybe they used their experience

The goal post has always been to replace human intelligence. I don't see any AI building Starcraft-playing AI's, or discussing how long it will be before they are replaced by even better AIs.

Uh no. Nice attempt at rewriting history, but there have been clear intermediate goalposts that people like you keep walking back on. Chess. Go. Poker. Starcraft. The list keeps building.

So what if you don't see AI building Starcraft-playing AIs? Your argument becomes more and more ridiculous. You claim there was absolutely no progress being made. And here, you reveal that your criteria of "progress" is replacing human intelligence. You realize you actually just shifted goalposts mid discussion, right? Now you're claiming any progress is no progress unless it replaces humans in one go.

Because humans have muscles that take time to move?

So? AlphaStar has its own limitations too.

Bots Don't Have to Suck Anymore

[mentil]

Playing a multiplayer game with bots used to be seen as an inferior experience to playing with real humans. Now imagine that instead of something like AlphaStar's utility function being set to trying to win, it's set to trying to make the human opponents have the most fun. Of course it'd need some understanding of the mindset of the player; they might not want to always win, or always have close matches, or possibly they're a sore loser. However, this could be inferred somewhat by player behavior (even outside of the match proper, e.g. in menus).

Put that in a game and ship it, and that could be a killer feature. People might prefer to play with a bot that'll guarantee a fun time, over a human that might rage quit or be an unfair match that leads to a one-sided game.
#MakeGamesSinglePlayerAgain

Re:Bots Don't Have to Suck Anymore

[Kjella]

This is probably much harder than you think, I sometimes play chess against the computer and even though it can match rating with human players the play is quite different. Like instead of making "reasonable" mistakes and miscalculations it's making random unmotivated moves that score poorly. Likewise, it rarely has any idea what a would be a good trap for humans, instead it's just surprisingly shallow at times as if it's maxed out the ply depth for that rating. And that's in a game that has so incredibly little nuance like chess, I can't even imagine the complexity of trying to act like a plausible rookie in Starcraft II.

Re:Bots Don't Have to Suck Anymore

[mentil]

For a turn-based game, particularly when you can see what the artificial stupidity is doing, you're right. In a realtime game like Starcraft 2 there's so much that could be done, that doing nothing with a particular thing at a given moment is a reasonable move. If you want to make the suspension of disbelief better, you can teach it what mistakes humans make, and allow it to repeat them when desired. AFAIK teaching a deep-learning algorithm how to fail like a human has received little research.

Re:Still Cheating

[Njovich]

The goal post has always been to replace human intelligence

The goalpost in AI research has always been chess. After that was solved some made random other goals, but there is certainly no concensus there.

Not sure what you mean by replacing human intelligence, it sounds like you are not happy until all humans are dead. Hopefully you are not an AI saying that. In the context of starcraft 2, this system has replaced human AI. If you want an AI system that can do everything in all fields that humans can do, well, it's not going to happen in our lifetimes, possibly never, and it's certainly not the definition of AI that AI researchers use.