Artificial Intelligence: Not just idiots anymore
Created | Updated Jan 28, 2002
Both the hottest area of computer research and one of mankind's oldest questions, artificial intelligence can't even be defined without someone, somewhere disagreeing. :-)
For the purpose of this entry, I will define Artificial Intelligence as the ability of an artificial device to achieve correct decisions from a limited data set, or just to fool people into thinking it is intelligent.
It could be argued that a certain computer program already in existence is artificially intelligent, but any such claim must be carefully considered. Humans (that is me, and hopefully you) have a tendency to anthropomorphisize. ><>Babelfish translation: we think animals, plants and inanimate objects act/look like humans.
Whataloadarubbish. The Turing Test suits me fine; when a computer chats with a human, and the human thinks it's chatting to anither human, we're there. For chatting programs. For other programs, similar tests can be devised. The central feature is:
Something has AI when we can't tell the difference between it and a human.
Several techniques are currently being used in AI research:
-Case Based Reasoning, or CBR:
Computers are good at logic, so we should be able to get them to think by logic. CBR uses ridiculously large numbers of 'rules' to try and get a computer to reason things out. Current level of advancement: Pathetic, except for expert (specialised) systems.
-Neural Networks, or NN:
The brain is made out of countless little cells, and we can make artificial brains out of artificial (usually simulated) brain cells, or 'neurodes'. It works, but unfortunately, just like a real brain, you have to teach it. But unlike a real brain, you can't just tell it stuff. It can't learn by itself. At best, you have to punish and reward it according to performance. At worst, it's like performing brain surgery with a toothpick and a soup spoon, and no idea how the brain works. Current level of advancement: Good at things like picture recognition, but still not up to human standards.
-Artificial Life:
The idea is that you can't just create an AI out of nowhere. It's got to have a real life of its own. So you simulate an artificial environment and breed Virtual Animals with a Genetic Algorithm. Can be combined with other AI techniques. Current level of advancement: Goldfish simulation, possible strange goblin-like creatures as well.
I've missed a lot of stuff out, and I'm sure a lot of people could contribute to this article. So please do, but nothing like 'AI sucks so foret that article'. Constructive criticism only please. :-)
For the purpose of this entry, I will define Artificial Intelligence as the ability of an artificial device to achieve correct decisions from a limited data set, or just to fool people into thinking it is intelligent.
It could be argued that a certain computer program already in existence is artificially intelligent, but any such claim must be carefully considered. Humans (that is me, and hopefully you) have a tendency to anthropomorphisize. ><>Babelfish translation: we think animals, plants and inanimate objects act/look like humans.
Whataloadarubbish. The Turing Test suits me fine; when a computer chats with a human, and the human thinks it's chatting to anither human, we're there. For chatting programs. For other programs, similar tests can be devised. The central feature is:
Something has AI when we can't tell the difference between it and a human.
Several techniques are currently being used in AI research:
-Case Based Reasoning, or CBR:
Computers are good at logic, so we should be able to get them to think by logic. CBR uses ridiculously large numbers of 'rules' to try and get a computer to reason things out. Current level of advancement: Pathetic, except for expert (specialised) systems.
-Neural Networks, or NN:
The brain is made out of countless little cells, and we can make artificial brains out of artificial (usually simulated) brain cells, or 'neurodes'. It works, but unfortunately, just like a real brain, you have to teach it. But unlike a real brain, you can't just tell it stuff. It can't learn by itself. At best, you have to punish and reward it according to performance. At worst, it's like performing brain surgery with a toothpick and a soup spoon, and no idea how the brain works. Current level of advancement: Good at things like picture recognition, but still not up to human standards.
-Artificial Life:
The idea is that you can't just create an AI out of nowhere. It's got to have a real life of its own. So you simulate an artificial environment and breed Virtual Animals with a Genetic Algorithm. Can be combined with other AI techniques. Current level of advancement: Goldfish simulation, possible strange goblin-like creatures as well.
I've missed a lot of stuff out, and I'm sure a lot of people could contribute to this article. So please do, but nothing like 'AI sucks so foret that article'. Constructive criticism only please. :-)
