Item — HN Reader

Not meant as a criticism, just an observation: The novelty of these shapes hidden in images is already gone for me. It's amazing to think how fast the hype cycles around a clever concept will accelerate with generative AI being so available and easy.

Meanwhile if people still had to do this stuff by hand in Photoshop, you could probably milk the gimmick for at least a couple of years Thomas Kinkade style.

I'm reminded a bit of million dollar homepage and all the fast-follow clones that each earned about 86¢.

4 comments

There’s also a lack of actual communication. Yes, an AI image generator can stick a pattern into an otherwise totally different kind of image. But there’s no communicative meaning there other than what I just mentioned. There’s no cleverness, nothing interesting to discover in the details, no space where the imagination can fill things in because it’s all plainly arbitrary, no deeper meaning other than a shape embedded in a pattern fill, however complicated the fill might be.

I went looking at the details of the original spiral medieval town and all I found was a woman’s skirt standing on its own with no body. And I know perfectly well the AI isn’t trying to say anything clever with that.

That’s true. Even when AI learns to not put torso-less legs, the wonder will be gone.

Prior to AI, you’d stare at an image like this and I believe most of the magic was marveling at the cleverness & craft of the person who painstakingly created the illusion. That you could think, “wow I’d never have the patience, talent, or dedication to make something like that.”

IMO there's a place in the world for easy, low-cost niche services that save you time.

I use things like remove.bg (a background remover) or VectorMagic (for raster vectorization) frequently, even though I have decades of experience in Adobe. Sometimes it's just easier/good enough to go through a dedicated tool that repackages some common workflow into a simpler wizard rather than have to DIY everything.

I tried to set up my own stable diffusion workflow the other day, using the ready-built GUI app Draw Things. Several hours of work and many gigabytes later, I managed to create several iterations of low-quality images that reminded me of early AI a few years ago. None of it approached the ease of use or output quality of, say, Midjourney, much less the state of the art these days (whatever it may be).

Maybe I could've tinkered with it for a few more days/weeks, but why? It's just an occasional toy hobby use for me, not something I want to get a degree in just to understand.

My main gripe with these services is their pricing model. I don't want to subscribe to something I might only use a few times a year. Wish I could just plop down like $10-$20 for a bucket of tokens that I can use-as-I-go. But that's not very sustainable for their business, I guess.

I think that's just about the context of where you see the content.

It's not that interesting as a hacker news post, or somewhere else with a tech audience.

The actual value for these things is in the mainstream, making content creation and graphic design easier.

In those fields, it's not about the novelty of, "wow that's impressive, how did they do that?". It's just about creating engaging content which is aesthetic and pleasing to people.

I expect it to be less engaging/pleasing and more instant article-filler material. To actually make something artistically interesting still requires effort, just a different kind of effort. The vast majority of these sorts of images will probably be uninteresting gags and background art.

In a way, the same is true of photography and digital art already, so this isn't a total condemnation (that is, most pieces have almost no value). I just don't think this particular kind of image generation is going to be a vehicle for anything noteworthy.

> The novelty of these shapes hidden in images is already gone for me.

Yeah I think it took approximately 1 AI image of cats spelling naughty words for me to crest that hill and lose interest in the entire technology. It's a technique that was only ever interesting because it took effort and cleverness, it turns out.