Hacker News Reader

Comment by frognumber | Hacker News Reader

To be fair, I've had similar experiences with professional illustrators. I describe what I want, and... something comes back.

In practice, the technology will improve. I'm still waiting to see where the ceiling is to know how many people will keep their jobs. I see similar articles about codex, and v0 won't take my job, but v3 might. Or v3 might be identical to v0. It's too early to know.

What I have found all these tools to do is to make me use whatever they generate far more. There was a pretty high bar for custom illustrations before DALL-E. With DALL-E, my last presentation had a half-dozen custom images. It was awesome. I'd never hire an illustrator to do that, but I definitely did "hire" DALL-E.

codex and gpt-3 are starting to change how I work.

If I use illustrations 10-fold more, which I seem to be doing, will illustrators get more work? Less? I don't know.

It is a brave new world, though.

View on HN

I think you touch on a good point. I think, like the automation of other skilled labor, the first jobs to be affected are the low-hanging fruit jobs. The ones where they just need something good enough. While I think tools like these will decrease the total number of jobs, I think the jobs left over will be ones that artists will enjoy more, as they will require more creativity than just writing a prompt.

This wasnt the case with industrial revolution. Most craftsmen either ended up doing a lot more joyless work as cog in factory OR they had to accept/adopt the factory process moved their creativity to design and ultimately lost their craft.

Main users of these tools will have to be illustrators. It will drive their prices down (it will be race to the bottom when you compete with instant AI) and over time there will be fewer and fewer people who will be able to create these visuals without AI because there will be no incentive to learn. Everyone will become prompt operator expert that will be the job.

You can imagine the slowly dying generation of illustrators in 40 years doing interviews for local TV about their wierd craft called drawing. (just like now you can see docs about scottish grandmas making tweed by hand).

I wonder where will the people move up from the knowledge work. One would hope it would be more free time but we all know how that worked after industrial revolution :)

This is an excellent point and easy for designers themselves to overlook. They may see DALLE as a tool that some are claiming does their entire job, and it can't quite yet. But it's really a question of how close it can match the actual result of employing an illustrator.

Yeah a lot of these articles are basically “it couldn’t read my mind so it’s bad”. I’ve employed a few artists before and the workflow is quite similar to the version 0.1 we’re dealing with for AI art now.

Explain your concept and then get 1 out of 10 illustrations kind of in the right direction, start iterating on that for a few generations and so on.

A flood of illustrations may also devalue illustrations themselves.

Often people value things specifically because they're scarce, and because they took effort to make (e.g. hand-made vs factory-made).

So I wonder if AI art will devalue all illustrations to the level of memes and stock photos. Or maybe some illustration style that AI is bad at will emerge, and become more desirable?

Images are already devalued on the web. Most people think that an article needs an image to not be a “wall of text,” with the result that many articles have barely relevant images that don’t add anything to the article at all. Like hotel room paintings, however nice they are, they are there to take up the space because an image is expected.

Actually relevant images are real photos of the thing you’re talking about, charts and graphs, screenshots, diagrams, and so on. Illustrations can be important content (consider a bird field guide), but often they aren’t.

Like the clip art and stock photos they will replace, these AI generated images aren’t a good substitute for the real thing, but I’m hopeful that at least people will have fun with them.

It might be interesting to consider how AI image generation could be used for nonfiction. Could you give it five imperfect photos of a bird species and ask it to draw a good illustration? Or maybe auto-create a good infographic from a spreadsheet? You’d have to do something to keep it from making up data points, though.

Reply on news.ycombinator.com