Super Upgraded StarCraft AI Still Beat by Humans

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
It is possible for computer to beat humans in chess. It is just massive computation. Even though today no computer is powerful enough to calculate all the possibilities, nor can human brain do, and computer is faster on simple calculations than us.

It might be possible for computer to beat humans in Starcraft. The game world is limited and static. It is more complex than chess but it is still limited, and computer can act faster.

In my mind, it is not possible for computer to be self aware, no matter how powerful they are. You can always fake intelligence (so a computer acts like it is on its own), but you can't create true intelligence.
 
It is possible for computer to beat humans in chess. It is just massive computation. Even though today no computer is powerful enough to calculate all the possibilities, nor can human brain do, and computer is faster on simple calculations than us.

It might be possible for computer to beat humans in Starcraft. The game world is limited and static. It is more complex than chess but it is still limited, and computer can act faster.

In my mind, it is not possible for computer to be self aware, no matter how powerful they are. You can always fake intelligence (so a computer acts like it is on its own), but you can't create true intelligence.
 
In the collectors edition DVD the AI developer/programer said that SC II is not chating for the first time. It will act like human. So it will attempt to scout you before it acts.
 
In SC2 the insane AI actively cheats by getting extra resources. The only way to win is by cheesing/rushing. In terms of macro computers have better apm than the pros.
 
RTS games honestly never had and likely never will have a good AI.
Many games even go so far to "cheat" hoping to give the AI some sort
of chance of winning. This is usually in the form of resources and the fact the AI never really had fog of war. Still the AI is usually easily beaten despite this.
 
"brains still better than silicone"

Are we talking about computers or supermodels? The verdict is very different in the two cases XD
 
[citation][nom]bv90andy[/nom]What's this Starcraft I keep hearing about?[/citation]
Don't worry about it, its not available under your rock.
 
It seems like a catch 22. They'll never be able to make an AI good enough to beat the best humans unless it was made by the best humans. Unless it's a true artificial intelligence that has the ability to rationalize the absolute best strategy to play, it's just going to be following general (possibly not-so-smart) guidelines given by the programmers. Therein lies an inherent weakness that can be exploited.

Especially in RTS games it's not like there's 1 magic formula that will win all the time.
 
Random generator = new Random();
x = generator(6) * 1;

if (x==1)
{
do these things;
}
else if (x==2)
{
do these things:
}
else if (x==3)
{
do these things:
}
else if..
 
Problem is AI programming doesn't look like that, its not just 'do random stuff' its pattern recognition, parallel simulation (what ifs, something humans do all the time), etc. we *will* get there and if the AI's way of thinking is modeled on human thought processes then their thinking will be recognizable to us...
 
"The original StarCraft AI was a notoriously bad cheater. And when I say it cheated, I mean it CHEATED. We're talking AI zerg rushes to your base, predictably, without exploring to find you, after just 5 minutes of play. AI units on different, locked teams completely ignoring each other and working together to wipe out enemy units one at a time. AI individually micromanaging buildings and units simultaneously all over the field. And that's just on standard difficulty!"

In response to this I say that a properly programmed, certified AI will notice that it has the ability to "cheat" and see that as a justified advantage it has over its opponent. A purpose written AI will not see the battle as a 'game' as humans do, but as a fight for survival and will use every method possible to ensure it wont be destroyed.
Once we create an AI without the confines of a purpose driven program, such as a game, we will see the birth of Skynet. :)
 
[citation][nom]xero9200[/nom]Seriously? That makes absolutely no sense. Why does the fact that humans program the AI mean humans will always be able to beat it? In something like SC, it's about speed and quick simple decision making, both of which computers are better at. Your comment is like saying humans built cars, therefore humans will always be faster than cars.[/citation]

In something like SC, maybe. But that fact is that we can not program an AI to be smarter than we are, unless your definition of smart is simply adding numbers quickly. Things like facial recognition, speech recognition, even actual conversations are massively hard to program. Things like the matrix/terminator etc are just fantasy. Any program can only do what it was programmed to do, and if you know anything about programming you'd know that you can't just 'program it to think' so easily
 
[citation][nom]Zingam[/nom]Or you could have just said that the point is that the super advanced AI that is also self-aware has the ability to learn and adapt, the way humans do and improve itself. And because it will run on a super fast computer with unlimited storage and it will not be wasting time to watch porn or to check the SMSs of its girlfriend to see if she's cheating on it, it will also spent much more time learning than the average human. So in any case we are doomed.[/citation]

That isn't a flaw in logic, you guys are both talking about hypothetical things that don't exist. That's like saying your logic is wrong, because if I had magic I'd be able to destroy the AI with my wand. AI like that does not exist, and is not likely to exist, and unless it does then you can't use it as an argument.
 
[citation][nom]jaygee02[/nom]In something like SC, maybe. But that fact is that we can not program an AI to be smarter than we are, unless your definition of smart is simply adding numbers quickly. Things like facial recognition, speech recognition, even actual conversations are massively hard to program. Things like the matrix/terminator etc are just fantasy. Any program can only do what it was programmed to do, and if you know anything about programming you'd know that you can't just 'program it to think' so easily[/citation]

I sure hope you find comfort in drawing that conclusion. Just because it "isn't easy" doesn't meant it can't happen. (See: 60 mph, speed of sound, man in space, man on moon, etc..) Part of 'creativity' comes from the combination of several related factors at once... once a computer can do that, it will do it better than us as well, because it will be able to compare millions of related factors at once while a very smart human only compares a handful.
 
[citation][nom]academicEngineers[/nom]@hurfburfthis has nothing to do with hacking or years of experience, this is a highly specialized field, using stacked queues of simplistic algorithms to generate complex behavior pattern in the most efficient manner, your typical professional hacker/computer engineer would not even come close to this kind of sophistication in their whole career, the majority of the works remains theoretical, something that universities excel atthe difference between academic and professional is stark, the fact that most professionals would only ever use 5% of anything there were ever taught academically is telling[/citation]

I never mentioned hacking. I'm talking about professional engineers. That's, like, the opposite of hacking. Do you mean to say you think these students are better than the folks at IBM who developed Deep Blue? The fact is, professional > academics. Academics try to reduce everything to the one problem they're specialized in. Professionals integrate multiple solutions to create practically viable results.
 
Status
Not open for further replies.