When you think of art, you probably imagine a big national gallery, showing you the highlights of art history. From Michelangelo, through Rembrandt, Rubens, Goya, Manet, Degas all the way to Picasso, Mondriaan and the contemporary ones like Koons and Hirst. The big problem with this is that you're only exposed to what is seen as the top of art production.
Did you know that about 90 percent of art created in the Dutch Golden age is lost? It was sold at markets and adorned the walls of your average person at the time. What's left is what humanity deemed worthwhile enough to keep around. Or, more realistically, what the rich deemed was valuable. Some of the less prestigious production survived, but you don't see it in the institutions that serve us the masters, you find them at auction houses and antique fairs.
If you spend time looking at these paintings, you can see why they're not taken into the prestigious collections of the world. Most of them are middling at best and a lot are outright bad compared to the masters. Our age isn't any different, if you go to contemporary exhibits at institutions you'll see what is considered the vanguard of our artists, but if you go to a commercial gallery you'll see different type of works, that range from the amazing to the horrendous.
Computers and artists have a strange connection. While traditional fine art has toyed with them, creating works that explore their use for new creations, they've for the majority had an impact on commercial art. In that field speed is often a necessity and tools that help with this are quickly adopted to speed up processes and set quicker and quicker time frames for the delivery of assets.
AI for creating images grew from this part of the art world. This was how it was mostly introduced about a year ago, as a tool to create images for content creators. Now you could have a picture accompanying your article that coincides with the content, almost immediately.
People losing their commercial work were up in arms, and rightfully so. The created images were nowhere near the quality of a handcrafted piece. They were just quicker and somewhat cheaper.1 They were also quite often made with models trained on their art, without any compensation to them. This is clearly a legal and ethical minefield, which is probably why most companies have still steered away from extensive use of them.
But the tools were available to the public and this lead to an increase in creative output by prompt writers, creating images quickly based on their ideas. Some I've seen were interesting concepts, others were cool fads on social platforms. None of them were what I'd consider good art.
Art is hard to define, but in general I go with a definition like an expression of a thought in an aural or visual form.2 If something is art isn't an interesting question that way. If something is good art is the more interesting one. Since Plato we've been struggling with this question and the entire philosophical field of aesthetics deals with this. So in no way do I claim to be an authority, but I have seen a lot of work and have literally gone to school for this.
My thoughts on this are that there are three things I look for in art: technical proficiency of the medium, social commentary or relevance and an ideology on what makes good art. In certain forms one of these three will outweigh another, but in general these are the three objectives I see that make good art, though there's still a lot of subjectivity and nuance within those. AI art, for better or worse, can never truly do this. While it can definitely look technically amazing, it is not actually applying a technique consciously, but following a prompt input by a human. You could look at the prompts as technical mastery, but then we should read the prompts standalone to decide if they are good and show all three objectives, in which case it will always lose out to man-made poetry.
Then a look at the social commentary or relevance. I've seen political works made with AI, but there's nothing there that couldn't have been done with Photoshop, and with a direct hand of a human leading the visual would've lead to possibly better and more pointed commentaries. The fact is that outside of the prompt, the results are arbitrary, based on what the AI can pull out of their database. In that sense the prompt works as a search to combine the right sources, which leads to the happenstance of an image that connects to the message the prompt writer is looking for. Contrast this to a painted masterpiece, where every brushstroke is meant to help convey the meaning. AI can at best get close to what the creator was imagining, but it can definitely not surpass the intention or add another visual layer of meaning by choice. It's limited by its database and the keystroke mastery of the prompt writer. It can't surpass those limits and add additional meaning intentionally.
Then the ideology question. This is where computers always falter. AI, in any form, is by limits or design, not able to provide an ideology. The concept of how things should potentialy be, not based on numbers but the concept of human morality, is something a non-human simply can't master. Even if there are intelligent forms of life that aren't human, such as self-aware AI would be, it is not human and therefore not able to truly understand the human condition. It can approximate and empathise, like humans can with others that did not grow up or experience their life, but it can never truly be a part of that experience. To live it is to understand it in a deeper way.
This means, within the arts, an ideology on aesthetics. An ideology on what is the right way and the wrong way to make good art. To create a category to be judged in. Abstract art is definitely not within everyone's taste, but there are certainly criteria that you must meet to be considered good art within that category. You can prefer Mondriaan's early work, where he was figurative, but you can't deny that his later abstract works are masterpieces within their concept of good art. They have created a concept and executed it.
AI is unable to formulate a concept on their own of what they consider good art. This is not even put into the prompt, but decided by the prompt writer by discerning the output and deciding what they like. But this means, due to the limitations of data mentioned before, that AI is only able to mimic other styles, to be a derivative and then judged on if the derivative is acceptable to the prompt writer. Sure, the prompt writer could have aesthetic concepts that they want to achieve with their AI art, but the issue here is that it can only be a remix of existing concepts at best. There is no chance of innovative AI art, because AI can't exceed the limitations of its programming. It can learn a style from enough data, but it can't formulate what makes it good or beautiful. It can simply take and approximate and then let the prompt writer judge. An accidental new style is possible, but this would at no point be a conscious effort.
So yes, AI art is art. So is your three year old niece's drawing of a horsey. It does not make it good art, or worthy of a platform in the public sphere. Nor does it make it very interesting to discuss, except for technologists who tend to have not more than a basic grasp of what makes good art. To me, the limitations of AI, it's inability to tap into the human experience and translate it into a worthwhile visual or aural experience makes it impossible to be good art, because it lacks the consciousness and the self awareness that is required for good art, next to the technical proficiency. An AI artwork that comes near this is like a cloud that looks like a chariot:: we see meaning in it, but it was never a conscious creation.
FOOTNOTES
1 The subscription model these services use is actually probably more expensive in the long run than commissions or licensing images a piece. I have thoughts on our subscription world that are better suited to another piece.
2 For those thinking about poems, novels and other written forms, remember that letters are visual as well.