Monday, September 5, 2022

Art and AI, Part II

I wrote about this a while ago, which ramblings you may read here. I may have gotten slightly sidetracked into making fun of how far "AI research" is from what many people seem to imagine its end goal is, in that piece. Let's take a more art-centric tack, to oppose the AI-centric one.

Some AI generated art piece won a ribbon at a state fair somewhere, recently. It is, obviously, a ridiculous, twee, mess, because that's the sort of thing that wins ribbons at these things. Be that as it may, it is certainly true that this category of AI engines will shortly, if they do not already, generate things that are not particularly twee.

So what?

I promised art-centristry, so here we go. I maintain that art is defined not by how it is made, or what it is made of, but by what it does. It strikes us. It at least has the capacity to enlarge us. But more than that, it means.

A mountain strikes us, enlarges us, but it would be silly to categorize it as Art simply as it stands there. It does not mean in the sense that we generally consider Art to. It does not stand for anything, it is not a symbol, it carries no message and can carry no message (unless, perhaps, from God, in which case everything is Art and the word dissolves into nothing, so let us set that aside.) A mountain, sublime as it may be, represents only itself. Art strikes us, and we believe in some kind of intention in that strike. We believe in the author and we wonder at the author's intent, no matter how vigorously we would cleave to a deconstructionist theory.

This can be broken down further, and indeed we have seen it broken down. The idea that an object might have semiotics, that it might indicate something other than itself, that it might symbolize, is separate from the idea that this meaning is intentional, that it is authored.

The Dada artists did some experiments with automatic drawing, automatic writing, that kind of thing. Others have used randomness, or mechanical means, to make paintings or whatever. These things can symbolize, they can have a semiotics in a way that a mountain does not.

Indeed, you could argue that the automatic drawings acquire meaning in exactly the way that a photograph of a mountain does, in the process of selecting which one to keep and which ones to discard.

If, as I believe, photography is not really an act of creation, but one of selection, then perhaps both automatic drawing and AI generated art are essentially the same as photography. The details of the construction of the thing itself are left up to chance, or the machine; while guidance is applied beforehand, and selection afterwards. Perhaps that is enough?

Will the AI replace the photographer? It might. Will I be able to pump in 3 phone snaps of myself and say "make me a powerful and moving portrait with a key light, a fill light, a hair light, a kicker, and a pink gelled background that also depicts me with a warm and authentic smile as if I am looking not at a camera but at you" and have it do that? Maybe. Probably? Sad for photographers, I guess, but jobs go obsolete, it's a thing. There will probably be a gradually shrinking pool of cases in which the explicit control of actually taking the picture will be worth the extra effort of actually taking the picture.

Of course, there will also always be room for someone to take photos of actual things that actually happen, as records, as documents of the thing, of the event. Perhaps all those photos will be run through the AI to pretty 'em up before we ever see them.

Probably a market for raw material for the mill (see "phone snaps" in the portrait example.)

I do not believe that the photograph as itself will carry any substantive cultural weight. I have given up on the idea that somehow the actual realness of the frame is what matters to ordinary people; what matters is that it looks real, not that it is real, and the machines can certainly make it look real.

Is it art? Sure it's art. It's as art as a photo, as art as an automatic or random drawing. Better than the last two, it tends to actually look like something, so it's not a largely theoretical footnote in art history. It looks like something, and it can mean in the approriate way. What's not to like?


  1. "art" just isn't a useful word to use here for me, for the same reason when people talk about what god wants they may as well replace that word with random noises for all the specificity it has

    the issue for me is A: creativity and degrees thereof and B: relation to real world events etc

    You're going to get people who really do use their creativity to make clever and arresting images/bodies of work by exploiting the right prompts, and that's great! I want to see that AI-generated work. A lot of, if not almost all of what I've seen, however, evidences the creative power of the people who wrote the AI, and a lot of evidence of humankind's creative mediocrity elsewhere. An engineer can make a guitar-playing machine that does it better than a human, and people will rightly be amazed, but really, they're amazed by the engineer's ingenuity and actually, any song of X complicatedness will do.

    The other issue is that properly great bodies of photographic work often come from very immanent and culturally/historically significant scenarios, and except for the possibility of making deepfakes and basically bullshitting, I really don't think that AI will override the great power photography has to deal with what people walk through, observe and experience happening here and now, even if on an image-to-image basis people tend to just like narratives and not truth-verification. Unless it's a purely aesthetic project, part of the narrative is usually the people and events surrounding the images, whatever context that is, and people are going to find a difference in significance for the same reason that we get impressed by good, well-earned musicianship from flawed, fleshy guitarists, even if romantically we might want to deal in "pure" images

    1. That's the kind of sharp, thoughtful, comment I pay the big bucks for. Thank you.

    2. My invoice is overdue. Just sayin'

    3. This art/AI stuff is much too hard for us unpaid interns. I'm happy to leave it to the pros.

  2. What if you described a scene to another photographer or painter or graphic artist and they produced an image for you. Would that be art? Aren't human beings just moist AIs?

    Robert Roaldi