AI art isn't for artists (third time's the charm?)

You know what the missing piece of the puzzle was? It was that AI art isn’t for artists.

Like, it can be. But for the most part it isn’t.

Okay so here’s a “thread” I wrote back in October, in the pre-ChatGPT era, right when everyone was freaking out about AI art, and I’m pulling it out for a third time to add some commentary on which of my thoughts have shifted and which I still think are relevant.

The quoted text is from October 2022, and the regular text is from May 2023. I think it might be interesting to revisit these thoughts (for the third time) because the technology has changed so much and become ubiquitous, whereas back when “AI art” was starting to become mainstream on Twitter, the quality was still terrible and it was mainly used for memes. ¹ ²

Like, when you think about it, if you need a bowl, you can commission a ceramic artist to make a beautiful one for you, right? Or you can make one yourself, which will give you varying results depending on how skilled of a potter you are. Or you can just go buy a bowl at Dollarama, because you need a bowl, don’t have (or don’t want to spend) a shit ton of money, and Dollarama gives you enough choice that it’s not really worth dealing with a potter or becoming one. It’s just a bowl, for god’s sake. Who cares?

I think the bowl metaphor is so much more relevant now. You can do something yourself, you can pay someone, or you can go to a store. The problem with using Al is that it’s equivalent to buying from a store that has zero quality control. You might get a damaged bowl. You might get a decent bowl-like object. You might get dirty plates instead."

I used to think that people would mostly use AI for simple things and grunt work, and boy, was I wrong. I mean, that is absolutely how some people are using Al right now, but I didn't have enough imagination to expect it to go much further than that. But of course it has. A few days ago I read this article about how a physics professor wrote a (somewhat simple, but still) paper end to end using Claude in two weeks and it's like, we can straight up replace mediocre grad students with this technology now and write a paper 20-50 times faster. You can't even compare it to buying from a store with zero quality control anymore. Instead It's more like having access to bespoke mass-production that gaslights you into thinking that your eclectic tastes are inferior now because maintaining them involves so much friction and is inefficient, and plus, why buy off-the-shelf when you can get *some* amount of tailoring for cheap?

So now you’ve got all these artists talking about how Al art is soulless or inexact or technically inaccurate or whatever, then laughing at tech bros who find it preposterous that anyone would spend years learning how to draw so that their random ideas can exist. And sure, if you don’t want to learn how to draw, you can commission someone. And so all these artists online are angry because AI art circumvents the need for commissions, and are saying that if people don’t want to pay someone they should get better at art. But for some people? It’s just a bowl, for god’s sake. Who cares?

I think it's kinda interesting how the tables have turned. As much as developers are probably those who benefit the most from this technology right now, they're also disproportionately affected by it. I've always said that I didn't want to be a developer, but that job is changing and I want to be what it's becoming even less. The bar for becoming a dev keeps getting higher and higher, big tech companies are starting to treat their devs like shit, and salaries are probably going to fall at some point. LLMs seem to be really good at dealing with structured and legible things. Tech companies have been trying to sell us the fiction that the world is a legible place for years now, with often disturbing results (that's a different can of worms). But hilariously enough, computer systems are pretty legible, so we get whatever the current situation is in the tech industry.

I think there’s always going to be room for commissions. But I think the reality is that some artists are going to be losing their jobs. The low level jobs? The ones where a client hires you because they really have no choice but they’re annoying and don’t give a shit about your expertise and you’re really not getting paid enough anyway? Those are all going to disappear within the next few years.

That’s just how technological innovation works. It just hurts more because the arts are so volatile already.

I think we’re starting to see this across the board, not just for artists. For example, I believe it’s IBM that announced that they will stop hiring for positions that they believe will be replaced by AI.

I think we all know that a bunch of the lower-level white collar jobs are actively disappearing right now.

I’m always going to laugh at people who type text prompts into AI engines and call themselves artists, because I wouldn’t buy a bowl at Dollarama and call myself a potter. But what I didn’t realize until now is that doing that is kinda missing the point.

So yeah, some of the AI art stuff is really wishy washy right now in terms of quality and there are a bunch of ethical concerns about where the training images come from. But it’s going to get better, and I don’t think it’s going anywhere.

I also think we should be forced to learn the basics of how chatgpt and others function at a high level. The ethical concerns are still just there though, which is a yikes.

I think at some point we're going to have to decide as a society whether we care about if people know things or can actually do things or not and I'm really concerned that the answer to that question is going to be “no”. We're rapidly getting to the point where doing your job yourself is seen as less competent than writing a prompt to get the AI to do it for you.

When I said better, by the way, I meant the quality of the images, not the ethics. It’s tech and capitalism, of course a bunch of it is going to be ethically questionable.

Here’s a last thought: when you want a picture of something, is your first instinct to hire a photographer? Or do you look for a stock image? And if you had access to a program that generated much more specific stock images, wouldn’t you use it? Sure, it’s never going to be as exact as if you could hire a photographer to put something together. But is that worth the time, money, and effort?

Or is it just another bowl?

ChatGPT is the peak site for acquiring “just another bowls” now.

RIP stock images.

Link to the original text. ↩︎
Link to the redux article. ↩︎