Live data from Hacker News

DeepMind and Blizzard Open StarCraft II as an AI Research Environment

deepmind.com

21–30 of 283 posts

Re: DeepMind and Blizzard Open StarCraft II as an AI Research Environment

#21
post #19
post #10

That's surprising. I thought Bliz didn't want anyone near sc2 but approved of sc1 being used for this purpose.

SC1 really doesn't make sense for this, 80% of the skill is just keeping on top of the mindless but mechanically intensive stuff, which is trivial beyond trivial for an AI. SC2's automated away most of this (pretty much everything but production cycles), which makes it a better measure for AI vs human.

I agree that SC2 is much easier for a human player. However, top players still have to do quite a bit of micromanagement.

Re: DeepMind and Blizzard Open StarCraft II as an AI Research Environment

#22

When Watson won at Jeopardy, one of its prime advantages was the faster reaction time at pushing the buzzer. The fairness of that has already been hashed out elsewhere, but..... We already know that computers can have superior micro and beat humans at Starcraft through that(1). Is DeepMind going to win by giving themselves a micro advantage that is beyond what reasonable humans can do? (1) https://www.youtube.com/wat…

I wonder if limiting APM would be a simple way to make the AI's play more "human" and less exploit-y.

Re: DeepMind and Blizzard Open StarCraft II as an AI Research Environment

#23
post #6

Blizzard should put in an AI-assisted play mode where players are limited to X lines of code that can be launched with keyboard commands.

That would be quite interesting, having humans handle the macro while the AI focuses on the micro. I'm reminded of "Advanced Chess": https://en.wikipedia.org/wiki/Advanced_Chess

> Advanced Chess is a relatively new form of chess, wherein each human player uses a computer chess program to help him explore the possible results of candidate moves. The human players, despite this computer assistance, are still fully in control of what moves their "team" (of one human and one computer) makes.

Re: DeepMind and Blizzard Open StarCraft II as an AI Research Environment

#24
post #9

Earlier quoted context omitted.

Do you have a citation? As far as I know, existing AIs do not stand much of a chance again humans [1]. My understanding is that even if AIs can beat humans is small skirmishes due to superior micro, they lose so badly at macro that it simply doesn't matter. [1] http://spectrum.ieee.org/automaton/robotics/artificial-intel...

Yeah, you're right. Not sure what the guy is talking about.

> the guy is talking about

I don't want to be overly semantic or PC on HN, but just saying the GP may be female, judging by their name on profile. Being misgendered could be very offputting and discourages participation, so you may have wanted to say "this person" even if it doesn't sound as offhand as you'd have liked it to come across.

Re: DeepMind and Blizzard Open StarCraft II as an AI Research Environment

#25

Earlier quoted context omitted.

My understanding is that in a full match, AIs still have no hope against humans, since even though they can crush humans at micro, their macro is still abysmal [1]. I'm not aware of a match where any AI has beat a pro human player at Starcraft -- I'd be interested in learning otherwise! [1] http://spectrum.ieee.org/automaton/robotics/artificial-intel...

would you love to be proven wrong?

Of course.

Re: DeepMind and Blizzard Open StarCraft II as an AI Research Environment

#26
post #19
post #10

That's surprising. I thought Bliz didn't want anyone near sc2 but approved of sc1 being used for this purpose.

SC1 really doesn't make sense for this, 80% of the skill is just keeping on top of the mindless but mechanically intensive stuff, which is trivial beyond trivial for an AI. SC2's automated away most of this (pretty much everything but production cycles), which makes it a better measure for AI vs human.

A lot of the balance of SC2 will immediately be destroyed by a semi-competent AI. One of the races has especially powerful early tactics that are primarily limited by a human's inability to multitask (reapers, medivacs, and liberators). I honestly don't think it will be possible for a human to beat an AI that just focuses on those strategies.

Re: DeepMind and Blizzard Open StarCraft II as an AI Research Environment

#27

Earlier quoted context omitted.

Yeah, you're right. Not sure what the guy is talking about.

> the guy is talking about I don't want to be overly semantic or PC on HN, but just saying the GP may be female, judging by their name on profile. Being misgendered could be very offputting and discourages participation, so you may have wanted to say "this person" even if it doesn't sound as offhand as you'd have liked it to come across.

In the NorthEast US, guy is often a gender neutral term. It's totally fine to say, "Hey you guys" to a group of only women.

Re: DeepMind and Blizzard Open StarCraft II as an AI Research Environment

#28
post #2

Any predictions for how long it will take for an AI to win against the world's best player?

AI can already win against the world's best player now. It's just a question of whether they can win with human-level micro.

Cool Starcraft related username!

Re: DeepMind and Blizzard Open StarCraft II as an AI Research Environment

#29
post #17

Related : Today I learned that a group of AI researchers has released a paper called: STARDATA: A StarCraft AI Research Dataset . According to one of the authors: "We're releasing a dataset of 65k StarCraft: Brood War games, 1.5b frames, 500m actions, 400GB of data. Check it out!" > Article: https://arxiv.org/abs/1708.02139 > Github: https://github.com/TorchCraft/StarData

The great thing about this is that it includes the game state throughout the game. It's been pretty easy to find lots of Starcraft replays, but the replays only include enough information to recreate the game (basically just the player actions). If you wanted to know what was happening in the game at the time the player made an action, you had to load up Starcraft and simulate out the game until that point. This dataset has already run the game for you and provided the data!
Post reply on HN