[ktRolster]

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/watch?v=IKVFZ28ybQs as one example

[obastani]

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...

[apetresc]

That's because there hasn't been too much concentrated effort on this problem yet, since you'd have to spend quite a bit of effort just integrating with the game engine.

Certainly a lot less research has been done on computer SC2 than computer go, and nobody expected a pro to be beaten there 1.5 years ago, either.

[TulliusCicero]

It's not that their macro is abysmal (macro in Starcraft refers to the mechanics of managing production and economy), it's that their strategy and tactics are real bad.

[flamedoge]

would you love to be proven wrong?

[sidusknight]

Of course.

[mattnewton]

That example might be misleading because I assume the AI has perfect information- I don't know how it could know which zergling was targeted before the tank fire landed without knowledge of the game's internal state.

In any case I saw in the comments above they are planning on limiting the APM. But right now they're not at the stage where they can compete with the in-game rules based AI, so it may be a little while.

[Waterluvian]

Thanks for that video. That's exactly what I hope to see. AI vs. AI with insane micro capabilities. I want to see SC2 played as close to a "perfect" game as possible.

[ktRolster]

Yes it would be amazing to watch

[sidusknight]

Alternatively, http://www.youtube.com/watch?v=DXUOWXidcY0&t=0m50s

[jahabrewer]

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

[ktRolster]

Limiting APM is definitely a step in the right direction, but there are ways to have super-human reaction times, beyond what a human can do, even while limiting APM.

So if we watch a match and see things that no human could physically do, we will know that the machine didn't win because of intelligence.

It would still be great, it just would be a simplification of the problem.

[chii]

What if the ai machine predicted very accurately what their opponents would do? Does that count?