[hughprime]

Oh, I think we're using the word "hard" in two different ways. I'm sure it would be very hard for me (or anyone else) to do, I'm just saying that it would be many orders of magnitude easier than building a true AI. Furthermore I don't think that one would be much of a step on the way to the other, since I really suspect that building a true AI is best done via either:

a) Reverse-engineering a human brain at the neural level, or b) Some kind of evolution

whereas I'm reasonably sure that the Loebner Prize could, given enough money, brains and time, be won by a large bag of special-case tricks based on careful observation of how the judges tend to behave in practice.