My Position On AI Art

Note

For context, this initially started out as a Twitter thread, but I’ve since decided to not spam the peeps who just follow me for YouTube storytelling stuff about this. But I still wanted to have this available publicly, so that I can directly link people to this which is probably the fullest description of my stance on AI art as of the time of writing - a stance which may change over time.

Introduction

I have a pretty complex opinion about AI art.

To start with, I’m very much pro-artist. That’s a scary first line to write, because any artist would assume I’m about to launch into something ultra pro AI, and usually people would throw hate comments straight away.

I have friends who are artists and care deeply about them and their ability to have a sustainable career and continue super great work. The amount of skills it takes to become a really good artist of any kind is absurd and should be respected.

And I think artists should be paid fairly and should be compensated well for their work. I do feel sad hearing when artists are done dirty by companies who fire them because AI is so powerful.

But this is where things start to get complex. I feel there’s two major functions of why artists do art: 1. art as a mode of creative and emotional expression, and 2. art as a functional unit of commercialisation. Whilst in reality these two blend, I think they should be logically separated in order to think about them better.

PART 1: ART AND AI ART AS CREATIVE EXPRESSION

For art as a mode of creative expression, I feel like expression is purely individualistic.

Humans learnt how to make art in a fairly lossy and unique way, taking inspiration from whatever and whoever they say, and having those visuals blended with all the bits that make up us. AI, on the other hand, does this learning in a rigid, mathematical, and precise way. But even if humans have a kind of imperfect and lossy mathematical equation compared to e.g. Stable Diffusion, every artist in the world borrows elements from other art.

Creative motivations exist separately from monetary ones. For example, a kid will find AI art easy to access and fascinating to play with. They won’t necessarily be thinking, “I’m gonna use this to make tonnes of money”. They will get super good at expressing art with AI, and where that fails, to learn art the traditional way. AI enables its own kind of expression, even if the “paintbrush” is abstract and digital.

It seems wrong to take away one’s expression, just because “that’s not the way it’s always been done, that’s not true art”. Gatekeeping anything sucks, especially when everyone has access to literally the same tools. There would be a stronger anti-AI argument to be had if access to the tools were unequal, but…they’re not.

Sometimes you see sentiments like “I can’t believe AI art got 100 likes and my hard work got 7 likes”. This is common, but sad. It implies the need for external validation to be an evaluation of how good a piece of art is, of how much someone is worth.

I get that AI art can be perceived as simply taking the work of other artists who didn’t give permission for their art to be used. Yet, that’s not unalike how humans are and how humans work when they themselves create art. We don’t maybe have equations in our brain that are as homogenous or well defined as a diffusion model…but there’s some commonality in our ability to look and attempt lots of creative pieces, and refine our art to look better.

Otherwise by definition, one would have no inspirations from which to draw from, and no skills they could have learnt from others. Everything is a derivative of what came before, combined into something meaningful. The paintbrush looks very different, but isn’t necessarily soulless.

But if AI art generates things instantly and to incredible detail, does it take away from the value of human art?

…my profile picture was handmade from a friend for my birthday. It’s special, because she clearly put significant effort into it. It’s not a super complex drawing. Even if AI art has loads of fancy strokes and lines and compositions and colours, this piece of art is still extremely valuable to me. When I was gifted it, it made my day. I don’t think the existence of AI changes that, because I don’t think AI’s existence limits creative expression or weakens the value of it.

So that’s part 1.

PART 2: ART AND AI ART FROM A COMMERCIAL PERSPECTIVE

Artists, like anyone else, should be rewarded fairly for their work.

It’s really unfortunate that it seems far too complex to pay all the artists whose work contributed to the training of an AI. To be practical: for any given piece, even if there was a gigantic pool of money to be shared amongst artists…how would you split it? For example, if a piece of art was given a financial reward based on similarity to other people’s art, how would that similarity % be determined? At both a human judgement and technical level, that seems to be an extraordinarily difficult problem.

Furthermore, if the artists who the art was trained on took inspiration from other artists…then does the money go to the original artist, and bypass the person who took inspiration from them?

“Then just don’t use AI art,” one might say, “it steals from artists and there’s no feasible way to pay them back.”

But the thing is…I think AI art actually creates really good opportunities from artists from a commercial perspective.

The reason why I’ve emphasised this separation of “as a mode of creative expression” and “as a functional unit of commercialisation” is precisely this.

Everything that exists in commerce simply aims to fulfil someone’s needs. Emotional needs is the biggest and most important subset of that, because generally speaking, emotional drives are pro-individual-survival.

Art does have a significant purpose in people’s lives. Whether that’s through entertainment or education, art is an extremely useful commercial tool in assisting the fulfilment of those needs.

But not all art that is commercially desirable is art that artists want to do.

I think a great example is in-betweening. So, in animation, there are keyframes which define the big and important parts of that animation visually. However, if you just had those frames, the animation would be super choppy. So therefore, (often more junior) artists are set to the task of in-betweening, which is to draw the frames in-between those keyframes.

Yet, animation is sometimes outsourced to studios in cheaper countries, in an attempt to get costs down.

Ultimately, the animation itself is a product, albeit a really expensive one to produce. I think if a commercial entity makes the non-creative non-interesting work much cheaper, then it allows artists to do something much more fulfilling and get properly paid for it.

There’s a cynical perspective - that actually might be pretty reasonable of - “the studio won’t pass it on to the artist, but instead will just rake in more profit and work the artists just as hard for cheap rates”. That will certainly be true for some studios.

But, I do think if a studio has a really healthy revenue and profit margin - from increased productivity gains by AI - they at least have the chance to pay artists more.

I think that AI is a tool that can help to achieve that aim significantly, by increasing the overall ability of a studio to produce content drastically. After all, entertainment from a commercial perspective is kind of the (digital) production of happiness, or whatever emotional need you want to fulfil. The use of AI in my eyes seems like a very pro-artist move, from my perspective.

When you type in a prompt into MidJourney or Stable Diffusion, or when you use ControlNet to make AI be controlled better by poses, it’s not like you’re sending an electric shock that’s sent to the artist whose art is in the training data.

Yet, making a studio more successful seems to be a very tangible win-win situation for everybody. There is a replacement of unnecessary work. Artists want to work on great work that moves people. Artists would be able to get paid more, because the studio would be earning more since they’re producing more. Of course, this does rely on the founders/management of the studio having their heart in the right place, which may be a big ask.

Conclusion

Regardless of how I phrase my thoughts and feelings on the subject, AI art as a topic elicits a strong emotional reaction to anyone who has thought about it before.

There are real examples of people who have been negatively affected by AI art in tangibly negative ways. For example, I dislike when people try to say that their AI art has been made by them by hand, and sell that art as if it’s not been made by AI at all.

I think that, when talking to artist friends, one of the biggest concerns is specifically this: the non-distinguishability of human-made art and AI art. After all, I doubt there’s a single person in the world who wanted to be tricked into thinking something is genuine when it’s not.

I’m also generally not a fan of much of how AI art is plugged as a community. YouTube videos that are like “How you can make $5000 with MidJourney in one day!” and then some dude holding a bunch of cash in the thumbnail…I mean, I don’t feel like that’s pro-artist either, it’s kind of just greed and a variation of the “make money online” niche.

There was one example of a case where someone took a screenshot of someone doing art of Hololive’s Ina, used AI to complete her back, then claimed that they were the original artist since they finished it first! That sort of thing is despicable.

But…I do think there’s a world where AI art isn’t just this thing that continues to negatively affect artists. If anything, I feel like it should be artists that benefit the most from AI art tools that they absolutely have access to.

Analogously, in programming, ChatGPT has revolutionised how software engineers program because they can write code like 10-20x faster. That code is likely trained on websites like Stack Overflow and Github, likely borrowing public code for training. But software engineers are generally delighted. Furthermore, it’s not even like they can even be replaced, because it’s actually non-trivially difficult to take a GPT-generated piece of code and make it into something that actually works in real life.

…

I think that AI art is actually pretty great overall. I remember the first day it was introduced to my life: I immediately thought, my god, I can finally get closer to making anime. I spent many hours studying to become a doctor and then later on being a doctor…so spending any of that time needed to get good at making art felt completely out of reach. I wasn’t thinking, “hey, I can use this tool to take someone’s job”. I’d just sometimes send friends AI-generated pics of their OCs, before the whole controversy really started cropping up, and it would make their day. No electric shock sent to the artist, just a bit more tangible happiness brought into the world that I could see with my two eyes.

It’s perhaps wrong to say I am pro-everything-AI-art related. It’s not without its dark side. I just hope there’s a day one can do something meaningful with AI art, without immediately being vilified for it.

But I hope to explain this rather complex and hopefully balanced thought process of why I believe, for the timebeing, I can be both pro-artist and pro-AI art. This tweet is hella risky for one’s reputation as a creator, but I do hope it makes some sense.

(PS: I think ethically sourced art like what Adobe is reportedly doing is pretty neat, and represents a more ideal solution.)