AI EI O
Old McITGuy had a server farm AI, EI, O (that's artificial Intelligence, Enterprise Integration and Oracle).
Ok, that was awful and no I am not drunk (yet). I was reading a good book on AI though on the possibility of intelligent computers. I have covered a bit of this in the Turing Test before in this blog. The book made quite a good point and one which I agree with.
The book was drawing the parallel between artificial flight and artificial intelligence. Basically early designs for flight were taken from birds which meant early flying machines were based on men with wings attached or machines that flapped. Neither worked for what are now well understood reasons (weight and volume blah blah blah). The point they make is that we compare AI to human intelligence and only in terms of a general interaction when it comes to the Turing test. I.e. the idea that if you cannot tell whether you are talking (typing into a terminal) to a computer or another human responding on the terminal then the computer would pass as intelligent.
The AI guys in the book say that it is the wrong target. Humans do some things very well and other things very badly. You can imagine the new IBM launch of its new super computer that could pass as human intelligence. Congratulations, you have now got a $1 billion dollar machine that cannot do long division in its head. But it can walk to the shops, get there and wonder what it came in for.
They have a point, so the focus is on improving the best bits about Human intelligence. No bird can fly at 45,000 feet for 10 hours. Inventing a bird like machine would have be genius but ultimately useless unless you need to land in trees and crap on people's heads.
But it struck me that the whole AI thing is doomed to failure for one main reason. I have always believed that we cannot define intelligence and therefore could never measure it. The book did not state that but the example they gave showed the issue. IBM's Deep Blue (I think that was its name) beat the Chess Grand Master but many critics state it was just a machine and was not intelligent. BUT, if my dog were to beat the Chess Grand Master, he would be considered an intelligent dog, no question.
The point is that we automatically have a 'yer but' filter. It does not matter what a machine does, it is a machine and it could do everything a human can do and do it a million times better and faster and people would still claim it not to be intelligent but just a machine. In other words many people find it hard to attach the label intelligent to anything artificial.
Until we get over that hurdle, we may never finish the race.
Ok, that was awful and no I am not drunk (yet). I was reading a good book on AI though on the possibility of intelligent computers. I have covered a bit of this in the Turing Test before in this blog. The book made quite a good point and one which I agree with.
The book was drawing the parallel between artificial flight and artificial intelligence. Basically early designs for flight were taken from birds which meant early flying machines were based on men with wings attached or machines that flapped. Neither worked for what are now well understood reasons (weight and volume blah blah blah). The point they make is that we compare AI to human intelligence and only in terms of a general interaction when it comes to the Turing test. I.e. the idea that if you cannot tell whether you are talking (typing into a terminal) to a computer or another human responding on the terminal then the computer would pass as intelligent.
The AI guys in the book say that it is the wrong target. Humans do some things very well and other things very badly. You can imagine the new IBM launch of its new super computer that could pass as human intelligence. Congratulations, you have now got a $1 billion dollar machine that cannot do long division in its head. But it can walk to the shops, get there and wonder what it came in for.
They have a point, so the focus is on improving the best bits about Human intelligence. No bird can fly at 45,000 feet for 10 hours. Inventing a bird like machine would have be genius but ultimately useless unless you need to land in trees and crap on people's heads.
But it struck me that the whole AI thing is doomed to failure for one main reason. I have always believed that we cannot define intelligence and therefore could never measure it. The book did not state that but the example they gave showed the issue. IBM's Deep Blue (I think that was its name) beat the Chess Grand Master but many critics state it was just a machine and was not intelligent. BUT, if my dog were to beat the Chess Grand Master, he would be considered an intelligent dog, no question.
The point is that we automatically have a 'yer but' filter. It does not matter what a machine does, it is a machine and it could do everything a human can do and do it a million times better and faster and people would still claim it not to be intelligent but just a machine. In other words many people find it hard to attach the label intelligent to anything artificial.
Until we get over that hurdle, we may never finish the race.
posted by So long and thanks for all the fish at
5:48 pm
0 Comments:
Post a Comment
Subscribe to Post Comments [Atom]
<< Home