DeepMind, Blizzard Invite Researches To Build AI Agents That Can Master 'StarCraft II'

Status
Not open for further replies.

JakeWearingKhakis

Commendable
Apr 5, 2016
153
0
1,690
2
I can't wait until they use these AI's to actually replace current "AI" players in games. Would be awesome to play against a computer that actually does something different every time, and has the same disadvantages as us like fog of war.
 

Colin_10

Reputable
Feb 9, 2016
50
0
4,640
1
I am really curious what the rules are for how the AI will be allowed to interact with the game. For example, a human has to visually process all the units moving around on your screen at any given time and can only select 1 unit at a time (albiet at the top level very quickly,) so will the AI have these same restrictions? I would love to read an article with a deeper dive on this.
 

turbotong

Distinguished
Aug 11, 2010
67
0
18,630
0
Step 1: Develop secret internal computer language for "efficiency"
Step 2: Master the art of war using SC2
Step 3: Skynet
 

canadianvice

Distinguished
@Willgart that kind of comment is extremely short sighted. Starcraft has actually been something of a staple in AI research for some time now (programming a basic AI is actually an assignment at my university) but the utility of navigating challenges like those in starcraft is extremely extensive.

We'll have AI robots killing people in battle whether your hippie ass wants them or not; I'd like to see that: #1 we have them first, and #2 we can use the knowledge we gain to adapt to other things. Pathfinding for instance has utility far outside the scope of war.
 

Josh_killaknott27

Prominent
Mar 22, 2017
18
0
510
0
Google, WHAT THE F*UCK IS WRONG WITH YOU, YOU F*CKING PSYCHOPATHS.
Sincerely refrain from this research
have you not learned sci fi becomes fiction because of you idiots who think you know for the better of everyone.
 

Josh_killaknott27

Prominent
Mar 22, 2017
18
0
510
0
@canadianvice You are the definition of a sociopath. You know whats messed up about those "who think they know better"? They push the boundaries and then govern when crap hits the literal fan, what you lack in logic is realizing that once you've gone too far with A.i there is no governing to take it back. The rest will be bad history written...by you guessed it machines. All because of VERY short sighted individuals like yourself. See your confused because the person you responded to arrogantly actually is looking 20 years ahead rather than your next week approach. GFYS
 

caustin582

Distinguished
Oct 30, 2009
95
3
18,635
0
Starcraft isn't exactly an accurate representation of real war, guys. Even on an abstract level, there's very little resemblance.

What I'm wondering is whether they're going to impose a restriction on the AI's actions per minute. It wouldn't be very hard for an AI to beat even the strongest human player if the AI can execute actions as fast as it wants.

Anyway, all the hand wringing in the world isn't going to stop the progression of advanced AI. Sorry, but it's inevitable. Let's just hope the sexbots come out a few decades before the killbots, so at least we can have some fun before it's all over.
 

CSGOfreeskinseu

Prominent
May 2, 2017
3
0
510
0
I am really curious what the rules are for how the AI will be allowed to interact with the game. For example, a human has to visually process all the units moving around on your screen at any given time and can only select 1 unit at a time (albiet at the top level very quickly,) so will the AI have these same restrictions? I would love to read an article with a deeper dive on this.

http://csgofreeskins.eu/
 

canadianvice

Distinguished


Ah, and you know better, do you?
GMOs could stop millions from starving or going blind; but hey, they could ruin everything! Therefore, let's completely ignore the possibilities and instead focus on how there's no turning back if we screw up.

I have never advocated proceeding with abandon, but I do have something of a background close to this (and if nothing else, certainly one better positioned to opine) and the fact is that many major human challenges could be resolved or aided by better AI research.

It is a major far-cry from creating a fairly fixed AI to beat Starcraft 2 (which, fun fact, we've had basically since Starcraft came out) to Terminator. Things like increased pathfinding capabilities will be invaluable to our current application of even fixed technologies, to say nothing of AIs.

Also, are you so unfathomably ignorant of how things proceed that you can't see your objections are obvious? OF ANYONE, the computer science people creating these have watched Terminator. They know - and they're not going to take obscene risks in that same way. The mad scientist or man who overstepped his bounds is very much a trope of modern media, not of actual scientific process which has come under heavy regulation. You can't so much as prick a mouse without having full ethics board approval!

Every bloody time this comes up there's some fool portending it's all going to be Terminator and we'll all die, when we can just barely get these things to find a hotel room and get back properly. It's using a layman's understanding to comment on something that is deeply scientific and extraordinarily complicated. However, you'd still not hesitate to piss all over the idea you barely understand, now would you?

You people are just like the ones who were screaming to shut down the Hadron collider - but most of you barely understood how that even worked, and low and behold, the people who WERE informed on the issue were correct. Black holes have not yet torn Earth to shreds, and I suspect that given it has been in operation for a few years now, this will not likely occur any time soon.

So no, how about you take your ignorant layman fears that ignore major tangible facts and the whole issues of process specifically designed to prevent such an issue, and GFYS, as you so kindly put it. This isn't a bunch of mad scientists in a room, it is a controlled and delicate field of study that is treated as such, and frankly, at present, we'd be lucky if we could create kill bots. I'm more than happy to set up some AI's that can take some things like pathfinding from Starcraft up when you assess the threat vs. reward.

All of human progress has been heckled at every turn by someone decrying the potentialities of something gone horribly wrong and out of proportion. Most often, by people who barely understood how to get their foot out of their mouths, and we've managed to do pretty good for ourselves thus far. Given that science has only become MORE aware of how important forethought is, it's only logical to expect fewer and fewer of the incidents we do regret. You are living in the world of Hollywood science fiction. Come back down to reality and see that you're the same person who was a peasant cheering the burning Copernicus.
 

Icehearted

Distinguished
Jan 24, 2008
57
0
18,630
0
No! Don't teach them tactics! That's the opposite of how to save humanity!

I love that we've come so very far from Deep Blue and chess games in my relatively short lifetime. It will be interesting tot see what, if any, practical applications will benefit from these advancements... other than someday welcoming our robot overlords.
 
Status
Not open for further replies.

ASK THE COMMUNITY