Can machines think?

I am interested in finding out how the responses are programmed within the computer. I mean are they responding to certain key words? Do they have certain 'beliefs' or 'responses' they defend to certain questions?
I think what I really am getting at is this: Can it really be called 'intelligence' if the computer is programmed to the say the responses it says? I think true intelligence for a computer would only occur if a said computer were to be able to gather information on a topic and critically form it's own opinions on the subject. I think if that did happen the results would resemble something of a Douglas Adams novel rather than a Asimov situation.

`…a machine seeks to fool a group of judges into thinking they're holding a text based conversation with a person in another room.`

One problem I can see with the Turing test is that under these conditions I can think of several people that probably wouldn`t convince other humans that they were human.

When we get to the stage whereby some machines and humans can pass this test, but other machines and humans can’t, we may have to rethink what we mean by intelligence and what the defining characteristics of being human are.

Modern AI methods tend to concentrate on programming intelligence, rather than responses. So the computer isn't "programmed to say" something, but "programmed to work out what" to say. Usually this software is informed by what we think the level of "programming" is in the human brain, so if we currently think humans have a certain level or programming - i.e. how memories and knowledge is stored - then the AI practitioners use that to design their AI software.
The difficulty is in moving from an "expert system" - some software that "knows" everything about medical diagnosis and answers with well-informed answers and shrewd judgement to a general knowledge that we would expect of a human being, or any intelligent life for that matter. Most humans wouldn't pass a very vigorous Turing test given the need for "informed opinions of a subject" and "crititally forming ideas from information available" - these are skills that not all humans have.
Plus the Turing test is human-centric, it wouldn't think much of dolphin intelligence and our best guess on that puts them as extremely sophisticated. The issue there being human comunication - a general intelligence test needs to be able to get over the problem of how to assess a being's intelligence without testing their ability to communicate with humans.

Thanks for explaining the programming that happens for modern AI wjlroe.
You raise an interesting point about the incapability of some people to pass a Turing test, but I think that if a test is calibrated to the appropriate background and age of a test subject it would be less of a problem.
I do however agree that the test is to based on human communication, but on the other hand what else should the test be based on? If a problem solving test were involved I think computers would probably be too good at solving them (unless they were programmed to make flaws and be less intelligent...in which case the computers aren't AI but AH...artificially human).

I look at it this way. We are bio, electro, mechanical machines. We exist, we make new versions of ourselves every few seconds of every day, so the building of what we call an "intelligent thing" is not impossible. The fact that we are here proves it.
So sometime in the near future somebody will find a way to recreate, simulate or mimic our reproduction mechanism, in part or whole to produce lets say a sentient machine capable of thought. Dont forget we have been at this game for hundreds of thousands of years, digital computers have only really been aound 70 of so years. Of course it may turn out that we will need something a little more complex and parallel that a basic Von Neuman architecture.
Personally I think neural networks are the key to this once we can build one that contains enough complexity. Simulating it software is a non starter, too slow, too linear to be of any use, which means we will have to build some special hardware, problaby bio, electrical, to get the density and parallel oprtation required.