This is exactly the problem humanodon and I ran up against in our initial PM conversation. And reading through the comments, looks like it's a common stumbling block for everybody. For almost every post, there's a different definition of what art is.

I arrived at my own definition in our conversation that I think both humanodon and I could get behind more or less, but in the end, any definition is going to run up against more or less arbitrary boundaries. Which, ironically enough, helps kind of answer the question, doesn't it? If art defies definition, if it's kind of a "know it when you see it" thing, how can it ultimately be taught? Which is I guess more or less the inverse of what you're getting at above:

if we agree upon a context for what we consider to be art... then we might say that to the extent that art can be communicated, it might be taught.

For the record, I think part of what makes art "art" and not technical writing or computer science or something is a confluence of intentional, directed movement with the presence, on some plane or another, of intuitive, a-logical design. If there's an easily defined formula to your art, then I'm not sure it's art, since art relies on a certain absence of logic. Which, leading to your question about robots and elephants- no, I'm not sure a robot can create art. But I'm less sure that the robot built to create art isn't an artwork in and of itself. Like the program that guy made to mimic classical composers. If the music itself is dictated by algorithm, then it's missing a key component of true art. But the program itself! Wow. Because it raises more questions than it answers, and it forces us to confront some uncomfortable implications about human creativity, and the nature of art. Wheels within wheels, man.

The elephant who paints? Who knows? Maybe, yeah.

That said, just because something is art doesn't mean it's good art. I can agree that anybody, including elephants, can paint without having to grant that all art is created equal. So if we allow for the presence of less effective art, it might soften the chore of defining art in the first place, and indeed offer some insight into what makes art what it is. If we know something is bad or good, and we can identify why, that implies something about the fundamental nature of the thing, doesn't it?

None of that answers anything. I don't know, This is why humanodon's question was so cool, though.