Artificial Intelligence

Discussion in 'Intelligence & Machines' started by kmguru, Apr 16, 2009.

Thread Status:
Not open for further replies.
  1. kmguru Staff Member

    Messages:
    11,757
    A lot of Computer Games companies are hiring Software Designers with Artificial Intelligence experience for Games Design.

    I am wondering if it is like document management that changed its name to Knowledge Management or what level AI is in the Games?

    Those who play a lot of such games, please tell us your experience. Does the Game AI interact with you (for you to say wow!) or it is just more good programming billed as AI?
     
    Last edited: Apr 16, 2009
  2. Google AdSense Guest Advertisement



    to hide all adverts.
  3. USS Exeter unamerican american Registered Senior Member

    Messages:
    2,482
    Its not that the AI thinks or is really 'intelligent,' it just comes pre-written with a finite amount of reactions to what the player does. In most basic terms, it can only allow certain inputs that will allow the output of the AI. Although some of the most recent AI is able to learn from the player. I'm not a really hardcore gamer, but this basically just stems from my experience from programming microchips and simple software.

    It would be nice to someday have an AI that can have some form of mechanized creativity.
     
  4. Google AdSense Guest Advertisement



    to hide all adverts.
  5. nietzschefan Thread Killer Valued Senior Member

    Messages:
    7,721
    Well I wish Creative Assembly would hire them. If I play one more total war game with easily flanked AI ....well I just might become a first person shooter, type of gamer

    Please Register or Log in to view the hidden image!

     
  6. Google AdSense Guest Advertisement



    to hide all adverts.
  7. kmguru Staff Member

    Messages:
    11,757
    The jobs are in Orlando, Florida...not sure who is there....
     
  8. kmguru Staff Member

    Messages:
    11,757
  9. kmguru Staff Member

    Messages:
    11,757
    “The Singularity Cometh” says Intel CTO Justin Rattner. Are You Ready?

    Within thirty years, we will have the technological means to create superhuman intelligence. Shortly after, the human era will be ended.—Vernor Vinge

    You are to be forgiven if you have yet to hear of the coming singularity. It’s a science fiction premise espoused by SF author Vernor Vinge back in 1993. Boy genius Ray Kurzweil put meat on the bones of the idea by writing multiple tomes on the topic. The premise of The Singularity is that soon, perhaps within one to four decades, we will be able to build machines with something rivaling human intelligence. Shortly after that happens, the age of humans will end as machines evolve like...well machines, and leave us to choke in their dust. Of course, that’s not how Vinge and Kurzweil see it. They’re optimistic that the machines will serve us. Or at least tolerate us. Apparently, they haven’t seen the Terminator or Matrix movies.

    http://www.edn.com/index.asp?layout=blogpostPrint&blog_post_id=610046061
     
  10. cosmictraveler Be kind to yourself always. Valued Senior Member

    Messages:
    33,264
    I use Chessmaster 10th edition and it teaches through an AI program in it how to make better moves if I want it to. I enjoy learning how to make better moves and this particular program does very well in doing just that. It can be very basic , if needed, or take you all the way to a masters thoughts of how to use certain moves and when. It interacts with me as I';m playing a game with it if I want it to. I enjoy learning from it as it is very constructive in how to do moves and why.

    Please Register or Log in to view the hidden image!

     
    Last edited: Jun 27, 2009
  11. Norsefire Salam Shalom Salom Registered Senior Member

    Messages:
    11,529
  12. thinking Banned Banned

    Messages:
    1,504
    artificial intelligence in the end will be our, Humanities, down fall in the end
     
  13. Nasor Valued Senior Member

    Messages:
    6,231
    The problem with complicated AIs in games is that if and when the AI starts doing something weird that you don't really want it to do, it can be very hard for the programmer to figure out what's causing the behavior or how to stop it.

    If you just program in a basic list of stuff for the computer to do like "if you run low on resources, send units to gather more from the nearest location," or "if your base is being attacked and you have troops more than X distance from your base, move those troops back to the base," then it will probably play more or less okay, even though it isn't going to spontaneously do anything cool or clever. But if you program an AI that uses a set of complicated weighted priorities to decide what to do on its own, it might do something cool and unexpected, or it might just do something weird.
     
  14. Nasor Valued Senior Member

    Messages:
    6,231
    There have been a few random games over the years that learned from the player. There was a space empire game called master of orion II where the players got to design their own ships from a huge number of choices of hull types, engines, weapons, etc. In the first version the computer would randomly try different designs and then note how they did when they went into combat, then pick the best one and make further variations of it. Although I believe they took that feature out and replaced it with mostly pre-designed ships in a later version, because while the computer was "learning" it would make all sorts of weird designs that didn't make any sense, and even after it learned how to make ships well they usually weren't as good as what the player could make with a little careful consideration.

    If you ever played version 1.0 of master or orion 2 and wondered why the computer kept building ships with nothing but 100 gyro destabilizers, that was probably why.
     
  15. kmguru Staff Member

    Messages:
    11,757
    I did not make a comment on the game AI...so here it goes.

    Having done some serious advanced programming in the past, I think, technically, you can not have a decent AI program without data history and a place to store computer generated data, rules, goals, strategies etc. that is continuously updated when program is running. The game programs just hold the data in temporary memory and when you stop...you lose the data. Perhaps someday...
     
  16. Cybernetics Registered Member

    Messages:
    89
    This is an intresting idea but I belive true AI in games has a fatal flaw:
    The AI would in the end come to a perfect stratigy making the game impposible to win, it would only work by limiting the AI's actiions which would damage the play value and so actualy updates from the games manufacturor are far more usefull thoughtI belive the Manufacturor could use it in the design of their Computer controled Oponents but not have it actualy learn in game.
     
  17. kmguru Staff Member

    Messages:
    11,757
    Not necessarily...you could have two AIs, one supporting the player and the other as the opponent with fewer resources like in the real life (USA vs. Iraq)....
     
  18. Cybernetics Registered Member

    Messages:
    89
    True but this would lead in RTS's to an AI with imediate responses and a perfectly constructed army and in FPS's a perfect Ai will have dead aim, perfect prodiciton and imposible dodging abilites. They would be able to do everything the player could but better thus their equipment would have to be vastly reduced or algorithms introduced to limit its capablites like random bullet deviance or variable unit damage.
     
  19. kmguru Staff Member

    Messages:
    11,757
    There is a game called "Command & Conquer - Kane's wrath" where the computer cheats. It has only one resource but builds assets faster and more than the rules allow for the player who has 4 resources.
     
  20. cluelusshusbund + Public Dilemma + Valued Senior Member

    Messages:
    7,985
    An if it was complex enuff it woud be indistinguishable from biological creativity.!!!
     
  21. Ganymede Valued Senior Member

    Messages:
    3,322
    Computer A.I has made leaps and bounds in video games. I've been an avid gamer all of my life, mastering to many titles to count. I can tell you this, in the old days the A.I had no sense of it's own surroundings. In first person shooter games the old school A.I would always leave it itself exposed and allow you to chop him down. Now, no matter where the A.I adversary is at, he'll use his surroundings to increase his chances of killing you. For instance, now the A.I will take cover behind cars, shoot any flammable or combustible object near you, set you up by throwing a hand grenade to bring you out of cover while his sniper buddy blows your head off the minute you expose yourself to flee the grenade. Also, I remember when I.B.M's Deep Blue beat Gary Kasparov in 1997. The first time a computer defeated the worlds best Human player. It was only a 200mhz computer, I assume Kasparov wouldn't stand a chance against todays computers. So that should answer your question.
     
  22. nietzschefan Thread Killer Valued Senior Member

    Messages:
    7,721
    120MHZ actually. 30 of them though.(processors)

    Frankly, I have yet to see an A.I better than 1993(?)'s Master of Orion. I think it COULD "cheat" (Production) at higher levels, but it was not necessary to play at that level unless the human was save/load cheating.

    You are talking about improved path finding and scripting rather than REAL A.I decision/behavior modifying with your FPS games.
     
  23. kriminal99 Registered Senior Member

    Messages:
    292
    It is good programming billed as AI. Tackling this issue can be done in mathematics even before programming. The question is what is the underlying algorithm actually doing. It's easy to tell the difference between actual strong intelligence algorithm and an algorithm that just tries to mimic what an intelligent being would say or do.

    The most advanced AI techniques deal with something in mathematics called the classification problem. The problem is to classify for example, and email as either spam or non-spam being given only the email. The classifier can have experienced other things in the past and have some memory of them however.

    All current approaches have a problem called overfit. This means that the classifier works well on examples it has seen before, and examples within range of what it has seen before. But horribly on novel examples.

    In practice this makes them useless. The advanced non-linear methods (multi-layer perceptrons for example) are given up in favor of computationally cheaper linear methods (PCA) which can recognize the simple cases where over-fit isn't a problem.

    To give an example, suppose a classifier must determine if a picture of a room is a picture of a classroom, having been trained on various pictures. If the training sample all have green chalkboards, the classifier will identify the green pixel grouping as a useful indicator. Then if a novel picture of a classroom depicts a black chalkboard, there could be a problem. Never mind a dry erase board or even a smart board.

    These approaches will never work, in order to succeed we must reverse our approach to the problem. A human being's approach might be to say something like "Does the room have something on which can be written?" <Something on which can be written> is a class, but not the one we are looking for. It's just a class that human beings generate and recognize for no other reason than that it exists. This allows them to overcome the problem of overfit.

    So as you can see, the correct approach to strong AI isn't to build a classifier to solve a specific problem, but rather to build a class generator.

    In other words a data mining algorithm that has some method of generalizing its results into classes. I have developed one, but the school where I attend contains mostly posers using the social network as a crutch and are afraid to look at it since they can't determine for themselves whether or not it is legit.
     
    Last edited: Nov 11, 2010
Thread Status:
Not open for further replies.

Share This Page