3 ms·
Hand-crafting generative code is the opposite of DALL-E. It's about affordances. A paint brush is a stupid simple object, but it offers an almost infinite rang
by TheOtherHobbes 4y ago
Hand-crafting generative code is the opposite of DALL-E.
It's about affordances. A paint brush is a stupid simple object, but it offers an almost infinite range of creative affordances, and almost endless potential to invent new ones.
A camera is also an affordance machine. You don't discover images, you frame them. And you interpret them by selecting and manipulating available light, tone, and colour.
And then very likely you do even more work later in Photoshop (etc.)
DALL-E has almost no affordances. Prompt goes in, studenty surreal "art" comes out.
There's no creative control. It's disposable novelty. That's all.
There are AI art systems that apply transformations to photo and text prompts. They're much more interesting, especially when they produce images with very high production values and genuine surprise.
So it's not that AI art can't be interesting. But DALL-E is what you get when people who don't care about that throw some ML tools into a bucket and pull out a blob of code.
It's not just that it's trivial, it's condescending in a way that useful art tools aren't.
- dahart 4y ago> There's no creative control. It's disposable novelty. That's all. People literally said this about the camera. I mean, I admit I’m playing just a touch of devil’s advocate here. As an artist on occasion, I’m not going to be wowed by the person who asks DALL-E for a landscape image. OTOH, I’m certain there can be creativity and skill put into the prompt text, based on time spent using DALL-E. And my main point is that the system is the interesting part, not the pictures. > It’s not just that it’s trivial, it’s condescending in a way that useful art tools aren’t. This opinion comes down to pure narrative though, right? If you see any given image and don’t know how it was made, there’s no way to judge it as somehow condescending. I’m certain DALL-E could spit out an image that you would not be able to identify as AI generated. If I didn’t have context, I’d have no idea the robot hand thing was DALL-E (and I have no idea if it spit out one of it’s training images, or if it mixed together multiple images.) You mention utility, and I don’t know what you mean by that. Could you elaborate on why you claim AI art like DALL-E is not “useful”? It’s pretty clear to me that DALL-E could generate clip art or ad art or stock photos quite easily, and would potentially replace humans that also pump out schlock.
- bambax 4y agoThis position is defensible, but it's only partly true. The author is foremost an editor of their own work. Take a lot of pictures, keep the good ones. Write a lot of junk. Keep the good parts, rewrite the others. Write a lot of songs. Publish the good ones. Etc. It's eminently possible that AI-art will allow faster iterations on ideas, and therefore help create more good art.