Um. No. Any reference to 'training' a computer implies the use of Neural networks, byesian statistical methods, or something similar.Spacewolf said:
And it really is training, NOT programming, because you give the computer a series of things you want it to be able to recognise, and the 'learning' algorithm involved tries to work out what's similar about them.
The key, and in fact the whole point of such training sessions, is that you then use the 'training' to let the program recognise things that it has never been exposed to before, but have similarities to it's training.
What's more, a lot of these techniques can't be studied very well. (The actual result of the 'training' isn't easy to work out)
A computer program can be analysed, and it's source code checked to see what it's doing, and from the code you should be able to work out what will happen with any given input.
But if you create a neural net, you can look at all the numbers and maths involved, and all the neural weights.
And, you can watch how these weights and bias values change as you 'train' the neural net for the task it's meant to do. (Neural nets are mainly used to 'recognise' things. It's what they're good at.)
But, it is almost impossible to work out what a neural net does, from the numbers that define it.
You have to work out what the neural net is intended to do from what kind of response it gives when you give it a specific kind of input.
By definition, this means it can do things you might not have intended it to do.
Whereas, with a typical computer program, you can, if you know what the code is, predict what the program will do, you cannot predict what a neural net will do by looking at the neural net itself. (Only it's behaviour can be studied, not it's 'design')
Working with these kind of AI routines can't really be called programming in a meaningful sense.
It's more abstract than if you were dealing with a human being, or an animal, but it still has much more in common with teaching, than programming.
