I think the whole talk about AI not being capable enough is bogus really; I'm with
@Sean on this. For the most part, it's already there, and insofar as it isn't, it's a matter of a very short amount of time until it gets there. Seems like presently it's mostly the step that needs to be made from "too perfect" towards "sufficiently wabi-sabi to trick us into believing it's real". That puts us in a position where photo-realistic AI is literally just that, and indistinguishable from the real thing.
Where would that leave us?
Well, let's put aside one important matter - when it comes to art, there's a heck of a lot more than photography. The thread asks about AI being used to produce art in a broad sense, so that inherently includes everything else except photography. This spans forms of art that do not need to mimic anything that already exists. There are entirely new avenues that are open for exploration. What's limiting us mostly at this point is the limits to our own imagination and ingenuity. I follow AI 'art' generation with a slanted eye and for the most part, it really is uncreative slop - renders of barely clothed women in violent "Lethal Weapon, Kill Bill" kind of situations (what the heck is wrong with people). Or when it's a little more 'out there', it remains stuck at rehashes of Star Wars and Dune type of imagery of fantasy landscapes or space stations. Mind you, this is all being done with exceeding technical prowess and the images as such are pretty mindboggling/stunning. Particularly creative or artistically compelling they are not, and that's not due to the use of AI. It's due to the fact that most of this 'successful' imagery is made basically by nerds with an interest in IT, and not by artist. Wait until capable hands (minds) start wielding this technology. We're in for a ride, for sure.
The more important issue I'd like to put to the fore is the in my view kind of hilarious argument about AI being 'not good enough' as a means to disqualify it from the domain of the arts. That's not the point, at all, in my view. That AI can be used to make art is not even up for debate in my opinion. See above; the fact that for now, it's mostly relatively uncreative minds using it and this resulting in fairly unartistic slop (including the 'crazy realism' video above) does not mean the potential isn't there. It's like arguing that cars are useless because once in a while someone drives into a tree with one. It's a logical fallacy to discredit the technology on the basis of a non-exhaustive set of counter-examples. Black swans don't exist until the day you see one, and all that. There's no doubt in my mind we'll be seeing plenty of black swans soon enough.
The real question is whether or not we'll accept that art as actual art. And I think, it depends here on the degree to which we recognize a human hand in it. In the end, I think the question what is or isn't art boils down to what
@thinkbrown said very early on in the thread; specifically this bit (I do not necessarily agree with what followed):
This limits the scope of what art is to work produced with involvement and arguably control by the human mind. I think in the end, that's really going to be the dividing line. And that automatically means that imagery (or sound, or whatever) that's made essentially by a digital agent with no direct control of a person over the outcome will never be accepted as art by a broader audience - with the 'audience' being defined as people who have an interest in art to begin with. That's an important caveat, because the majority of mankind in my opinion has no real interest in art. People like pretty things, overall, but that's a different matter.
The underlying question is why we would draw a dividing line between 'art' and 'not art' on the basis of human control. In my mind, that's simple - we want to be able to relate to it, that's all. We can't relate to a datacenter any more (and in fact, much less so, on average) than to a pet cat. It's just too different a beast. When it comes to Ai-generated work, one of the key problems is that a datacenter is just too...good. It yields a perfect result, even if we require it to be imperfect - then it'll be perfect in its imperfection. For whatever underlying psychological or evolutionary reason, we can't really cope with that as humans. Probably it's just too threatening in the end and we can't cope with that in the way we usually do.
Think about it - why do people read tabloids? For that matter, why do people read biographies of people like Einstein or Musk? A large part of it is to be able to spot all the imperfections, vices, character faults, misbehaviors and flaws exhibited by the rich and famous. Those people we all somehow measure up against, who are the benchmark of success in one way or another - one of the main ways of dealing with them and the notion that most of us will never really succeed in life the way these happy few have done (that's why they're few, after all), is to focus not just on their success, but also on their failure. Einstein married his niece - LOLWUT! And didn't that video of Musk smoking a joint during a podcast recording go viral? Sex! Drugs! Human weakness!
It's a lame piece of Hollywood trash in many ways, but in some ways, Bicentennial Man is an interesting movie. At what point does Mr Robot get accepted (American-style: in a legal sense) as a human being? That's right - the moment he 'failed' in a human way: by (arguably irrationally) opting for mortality.
It shows in this thread as well. The moment the notion of Ai
autonomously creating art arises, the response is essentially "it can't be, it's not supposed to be!" Arguments are all over the place - it's not good enough, it's
too good, it can't think, it can't feel etc. In short - it's not human enough. It's insufficiently like me, and if it's not like me, I won't accept it. In the end, that's the criterion, the way I see it. And I think there's nothing wrong with that. If anything, AI is going to help us understand a little better what "being human" involves. Maybe that's the greatest 'innovation' it'll bring.