AI is not for art
Disclaimer: I am not an artist. I am a developer. This post was written from my point of view on the topic.
Hot take:
AI cannot generate genuine Art. It just can’t. It can try, but it cannot create something the way a human can.
What is AI good at?
AI is quite good at assisting. It can recognise a common pattern and continue it.
This has worked quite well in the programming world. In my experience, tools like GitHub Copilot work great for writing common written code, like boilerplate. Copilot would give me suggestions and I could accept and tweak them to my liking. In my opinion a smooth experience. It made my work easier and faster.
If AI is so good, why does it suck for art?
In the text above I talk about how AI is good and helped me in writing software. So why should it suck for art?
The answer: control. When I code, Copilot gives me suggestions, which I can accept and tweak afterwords. This is not the case with generative image or video AI. When creating an image with, for example, Midjourney, you get a completed image. This makes tweaking the image harder.
AI art tools also don’t integrate into an artists’ workflow. When artist draw, they usually start with a sketch, on which they improve upon until they like how things look. Then they start coloring and finally shading. (This is of course different for each artist, but this is the basic gist, which I have picked up from watching artists work on the internet.)
AI skips all of this and goes straight into the “its finished” stage.
So if the issue is known, why not fix it?
Integrating the current AI tools into the workflow of artists is not possible. This is caused by the way that the AIs work. Since they are trained on finished images and not the strokes of the pencils made, it is not possible to offer Codepilot-like suggestions to artists.
And this is probably a good thing. When an artist is drawing, they usually have in mind what the result should look like. An AI trying to predict how a specific line should look like and then rendering a preview over the already drawn image seems impractical and distracting. For this reason specifically I think that AI may never be huge in demand with artists. Sure, there will be people doing it anyway, but most of the people generating AI art are not artists themselves and most of the time do not care about the specifics of art or the way its made.