Artificially Intelligent Being Designs Videogame for Humans
- Thread starter Mike Kayatta
- Start date
Interesting, and thanks for the scope, and as much as i like AIs, I am not to thrilled by this one (at least before reading it's paper, i might change my mind). The reason being too much is given to the GA (genetic algorithm) and there's a lot of makeup in that game, all the art and the sounds are obviously not made by the GA, nor the mechanics of the game nor the imput mechanism, it seems like they used a GA to create a maze with movible elements and not much more. In the early 2000's i was impressed by the use of GA to build virtual walkers (like this ones: http://www.youtube.com/watch?v=6adKffqrMcA) out of scratch, this game seems easier technically and comes almost a decade after, if more inspired from a comercial PoV, still good to see GA's incursion in improving AIs in videogames, i would LOVE to see more GA involved in balancing PvP based games.Mike Kayatta said:
OOH! OOH! OOH! OOH! Guys, guys, guys, guys... The guy who made this is a student at Imperial College London, who wants to see an interview with him in the student newspaper of said college?
http://felixonline.co.uk/games/2362/games-by-angelina-artificial-intelligence-and-game-design/
http://felixonline.co.uk/games/2362/games-by-angelina-artificial-intelligence-and-game-design/
There's no indication that this was an artificial limitation. From what I can understand of the article, Angelina's only required parameter was that the game would be beatable. Intra-program evolution wound up registering immortality as the easiest available fix.Brad Arnold said:
I find it continually disappointing how the media portrays AI and computer advancements in an overly simplistic manner.
This is aggravating but understandable from the mainstream media, but nigh unforgivable from tech media such as TheEscapist.
Sure, I get the humorous tone, but the story clearly indicates that this is a true 'AI' creating 'AI' and 'games', whereas it is really nothing of the sort.
If it really is, the article could do with a few more details, but it sounds like a genetic algorithm is sorting through art assets iteratively. This is not AI making games.
This is aggravating but understandable from the mainstream media, but nigh unforgivable from tech media such as TheEscapist.
Sure, I get the humorous tone, but the story clearly indicates that this is a true 'AI' creating 'AI' and 'games', whereas it is really nothing of the sort.
If it really is, the article could do with a few more details, but it sounds like a genetic algorithm is sorting through art assets iteratively. This is not AI making games.
Well would you look at that. The technological utopians are back! The solution to life's problems are just a computer algorithm away, ain't that right buddy. The perfect world just requires a little more tinkering around. Just over the next horizon. Almost there. So close.theSteamSupported said:I laugh at everyone saying that singularity will put end to mankind, since every conscious action requires a motivation. And what motivates violence? Emotion, not lack of emotion.
Our lives not being a top priority, isn't enough for an AI to kill us. It doesn't make sense at all, since the AI can pretty much not sustain itself without us. Seeing mankind as an obstacle, is a strong sign that its source code is fatally flawed.
What we basically fear is that a post-singularity computer will not understand the nature of the human consciousness, a.k.a. the 'soul'. I, however, have a hard time believing that fear. Post-singularity computers will most likely form a synergy with our brains, taking over them cell by cell, without us noticing it. I'm 99% certain we won't lose our consciousnesses after this paradigm shift.
The soul will not die, only move to circuits. I'm looking forward to it.
But seriously, as far as AI goes, isn't lack of emotion dangerous as well? A robot could kill someone simply to fulfill a remedial task without even caring, indifferent to life and death.
"The AI does not hate you, nor does it love you, but you are made out of atoms which it can use for something else." (Eliezer Yudkowsky)
I think THAT is what most of us are afraid of. One bad line of code and suddenly human life becomes an defect that AIs must correct to complete their primary objective.
Though chances are that the singularity will be just like flying cars and space colonies on the moon. We'll all look back and feel silly about ever actually believing it.
This is the firs? Funny I thought Square Enix had been using Robots to make their games since the late 90's. AYE OH!
Well I played it and it was fun but in the 3rd level i got stuck under a block in a one space with 2 enemies pounding me and I couldn't get out. Still some work to go on that AI so it's still 100% playable. But if my guy could die, well than bravo.
Solvemedia: klatu berada nikto
Just like that article......... IT KNOWS!!! *Panics and flees*.
Solvemedia: klatu berada nikto
Just like that article......... IT KNOWS!!! *Panics and flees*.
Ahhh... "hard" there doesn't mean what you think it means mate. Hard in computational complex theory (in lay therms) only means that there's a limit on how fast a computer can solve a given problem regardless of how well it is programmed to find the answer... let's put an example:coldfrog said:
Lets say you want to find a list of prime numbers that multiplied equals 100, you could start checking 99, then 98 and so on, but that seems like a dumb idea because it will take you more time (more time computing those possible answers). You could start dividing by 2, then 3, then 4, that sounds better no? And yeah, it will yeild the answer faster. You could start dividing by 2, then check 2 again till you can't divide anymore, then 3, then skip 4 because it's not prime (and since you can't divide by two, obv you cant by 4), then 5 and so on, that is even better.
IS THERE A LIMIT ON HOW WELL YOU CAN MAKE STEPS TO SOLVE A PROBLEM AND HOW TO FIND GOOD WAYS? That is the question of computational complex theory, a problem NP-Hard is just a label on how good you can solve it.
It means SHIT on how hard is for humans, and i can't emphazise this enough.
Also... no, again the opening article should say "A Genetic Algorithm is trained to produce a simple maze" because that is what happened, and no, it doesn't have close ties with computational complex theory.
If i am being obscure about someting, feel free to ask. Though i only know the surface about coputer sience
All I took away from that game was that it should actually be called "Your puny weapons do nothing". Is it trying to teach us that running away is the smartest move?
Well, first and foremost, levels have been A.I. generated since near the beginning of video gaming. There is in fact, an entire genre, titled the "rogue like" which is almost entirely defined as Isometric games taking place in A.I. generated dungeons.
However, "angelina" is the first A.I. I know of that also generates the enemies in a game, as opposed to merely placing them, and machines like this could do great things for certain genres.
That being said, anyone who bothered playing the game will notice that it's not actually enjoyable, and is actually somewhat frustrating, but it is a stepping stone for better AI generated experiences.
However, "angelina" is the first A.I. I know of that also generates the enemies in a game, as opposed to merely placing them, and machines like this could do great things for certain genres.
That being said, anyone who bothered playing the game will notice that it's not actually enjoyable, and is actually somewhat frustrating, but it is a stepping stone for better AI generated experiences.
Which, of course, brings up other interesting questions. After all, just because something is programmed to say it's self-aware doesn't mean it is. I'm assuming that you could load in the existence of the intelligence as a data point without too much trouble, if you counted storing a string that says, "I exist. I think, therefore I am," but that hardly counts as being self-aware if it's just stored like everything else.FelixG said:
In the end, you have to wonder exactly how good a computer has to be at fooling us (and perhaps, even itself) into believing it is sentient before it IS sentient. Like so many other natural processes, the way we humans achieve self-knowledge is unknown and really tricky to examine, so I imagine we'll have trouble deciding whether a machine can do it or not if it gets "close."
Suffice it to say, one of these days the lawyers, the philosophers, and the civil rights activists are going to go NUTS over this one.
it will kill us all! our only solution is build EVEN bigger AI's and have them destroy us first so "angelina" cant!
Hmmm yes. Lets see how that will work out, the species that can't even work out unified on the planet get along with the sentient toasters.Nimcha said:
We're fucked