intelligent systems

Rob Harris (rob@hbinternet.co.uk)
Mon, 29 Nov 1999 10:32:19 -0000

>OR: how would you describe the "intelligent system" used by use
>by a bacterium to achieve its "strictly defined goals"? "It doesn't
>have a central nervous system, so how can it have one?" Then
>how did we get one? Not because we needed to! Think about it.

As always with these kinds of debates, the problem lies in the premises. It seems we differ in our ideas of the meaning of "intelligence". The rule of thumb is - it doesn't mean anything, but you'd hardly call a static finite state machine "intelligent", so some dynamic characteristics are required. Ask yourself how you would decide if your machine has "intelligence". You'd have serious trouble - not because the data is hard to get, but that the concept of intelligence is an animal fitness function, not an objectively observable entity that you can measure. Alan turing knew this when he invented his test of intelligence - the imaginatively named Turing Test. He realised that "intelligence" could not be measured by anything other than a human (as it is a human "fitness function"). The test involves giving your AI to a human for evaluation - if the human thinks the program is intelligent, then it is. The test involves no empirical measurements at all - after all, what measurements would you take? These days, probably due entirely to marketing, everything with a hint of dynamic behaviour is named "intelligent" - and it may as well. After all - if you designed a self-sustaining and repairing, replicating, evolving, little machine out of atoms - you'd not refrain from calling it "intelligent", and rightly so !