The thing that's fascinating about watson, is that it will ONLY get smarter. The more data it takes in, the more examples it sees, the more it understands.
It's interesting watching Watson, especially in the NOVA documentary, where it demonstrates in a game how it can learn. After each question is answered, that information is passed to Watson in a text file. Thus, it learns from its mistakes and knowledge holes, and gradually learns the context of the questions. The only thing it can't do, is that it can't hear incorrect answers, which could help it narrow down it's confidence intervals (the example with Jennings, where it repeated his answer).
Watson is going to be HUGE for IBM, and it will pay off in droves. I can see the END of the phone menu systems that suck ass, and being able to basically ask Watson 3.0 a question, and actually have it respond to what you're saying. Hell, I could see Google getting in on the act, and having the same kind of functionality.
A tool like Watson would make the Google Books project explode into a whole new range of awesome. Given enough input, and enough interaction, in about 20 years, we might see it make a run at the Turing test.