I've not read much about this definition, but it seems that a problem with focusing on the ability to predict is that to get useful work done, you have to reduce everything to a decision problem. You have an ideal fitness function, but how are you generating ideas, plans or theories to evaluate?
I personally prefer (although it's certainly less measurable) Ben Goertzel's definition - the ability to achieve complex goals in complex environments. It's still wishy-washy enough for people to write off any given achievement of AI though, I guess.
The brain as a prediction and inference engine (the Bayesian brain) has been accepted by many in neuroscience. I think this article is novel because he rebels against this idea. I do think that many of the things that we consider intelligence are prediction, image recognition, speech comprehension, understanding other's theory of mind, and many games (e.g., chess).
I think Deutsch's argument is that there is also creativity (and creative reasoning) that is omitted from today's approaches to AI.
