The sole definition of intelligence is under heavy load of reconsideration the last decades with the emergence of a better knowledge of animal cognition, for example.
We like to think that AI is only statistical models, which may be true (is true?), but we have no idea that what happens in brains is anything else than a highly optimised, deconstructed, such model.
Current datasets need 10ks pics of a cat while a kid only need to see a cat twice? Good for the kid, but there's no hint there's a fundamental difference of process.
I just don't see anything intelligent yet. People somehow have gotten confused with the success of machine learning being treated as intelligence. We have a lot of statistical models of things that are very successful, but those aren't considered intelligent. For reason, machine learning telling you something about data has suddenly been treated as AI. Machine learning can do some impressive things, but I think its short-sighted to equate AI and machine learning.
Intelligence is really a tough thing. Watching a video of even a single-cell organism displays a sort of intelligence and behavior far beyond anything I've seen of machine learning. So why is it intelligent? Or is it intelligent? I'm not entirely sure, but my point is that machine learning is orders of magnitude incapable of describing (i.e., modeling) even the simplest self-directed and self-adapting behavior that we see in the real world.
/s?
I always thaught those were synonyms.