Do you think we will ever create a machine that can pass the Turing test?

Answer 1

Yes, but the test is about the "appearance" of being human. As far as artificial intelligence goes it will require a much more extensive test.

Answer 2

Heh. It depends a lot on who's doing the testing, on whom, and what they are after.

Take a look at dialogue 22 in link 1. That was generated with an agonizingly simple response program... and the person spent more than an hour communicating with it without detecting anything fishy. Of course, they weren't LOOKING for anything fishy, either, so it wouldn't be a valid 'Turing Test'.

It is likewise pretty easy to imagine a person trying to distinguish between a very clever computer and a very stupid human and having a great deal of difficulty. Some people, after all, aren't very creative or well-informed. You'd probably stand more a chance of giving your software away with perfect spelling than anything else! Still... who wants to be the inventor of a program that can replicate an idiot?

I seem to recall an interesting AI that was developed not too long ago that gave its own lecture on AI at a conference after booking its own rooms, plane tickets, and navigating all the way to the conference on its own - of course, it was armed with a certain amount of information beforehand. Can't find a good reference to that but found something just as well: a small robot in a new environment about which it had no information escaped from its enclosure and made its way out of the building it was in (link 2). I know people who have trouble with that sometimes!

So yes. It's really only a matter of time. Even if it just comes down to the most hardcore perfect simulation of a human brain derived completely from physics, chemistry, and genetic information. Only a matter of time.