Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Stable Diffusion is a big deal

simonwillison.net

Post reply on HN
Print view View original

51–60 of 488 posts

Re: Stable Diffusion is a big deal

#51

Post by nomel »

michaelchisari wrote: post #9
canjobear wrote: post #7

Earlier quoted context omitted.

> Second, we need the ability to persist a design. I See yesterday's Stable Diffusion article: https://news.ycombinator.com/item?id=32643564

I haven't had a chance to try it yet, I'm optimistic but skeptical that it can have true persistence like I'm referring to. Especially since it requires training your own model which requires multiple images of the same asset/character/object/etc.

For what I think you have in mind, I suspect it will eventually not be "image to image", but " to + image", for that guess to be remotely repeatable. That "ai thing" is probably the persistence you're talking about.

I think that thing will neccesarily contained representations of dimension, behavior (physics/bones), and "style". Without the "ai thing", if only using an image/text, the character would have to be impossibly represented in the model, so it could guess all of these things predictably. For example, what that character looks like from a side profile, or behind. What if it's an alien, and its arms should always bend backwards? Could a text representation ever be made to completely describe this, with good reproducibility? Probably not. But, I assume some non-human representation would have a better chance.

As is, if something known is required, I think the behavior of these models can be considered "destructive" to the input image, more often than not. For this reason, I think artists are safe, for the time being. :)

Re: Stable Diffusion is a big deal

#52

Post by chromakode »

betwixthewires wrote: post #27

One day, images generated with these AIs will become ubiquitous online, and new AIs training on data will wind up with feedback problems because they got trained on photos that are AI generated, possibly even ones they themselves generated! This tech is a huge deal. Huge. Ubiquitous unique, beautiful art. Entire movies made by machines. Too many to even watch them all. Actor models that aren't real people owned by th…

It reminds me a bit of "Low-background steel", which recently popped up again here. Images older than 2020 are predominantly human generated. Anything later is suspect!

Re: Stable Diffusion is a big deal

#53

Post by kfrzcode »

michaelchisari wrote: post #6

After using SD heavily for a week, I half agree with this. It is incredibly disruptive, and it's wild how much it accelerates the creative process. I'll give you that. But two things I've noticed: First, artists will still have a massive advantage over non-artists with this tool. A photographer who intimately knows the different lenses and cameras and industry terms will get to a representation of their idea much fas…

> artists will still have a massive advantage over non-artists with this tool.

Good; I hope the future of software engineering AI generative tooling is also more useful for software engineers.

Re: Stable Diffusion is a big deal

#54

Post by Morgawr »

ipaddr wrote: post #36
betwixthewires wrote: post #27

One day, images generated with these AIs will become ubiquitous online, and new AIs training on data will wind up with feedback problems because they got trained on photos that are AI generated, possibly even ones they themselves generated! This tech is a huge deal. Huge. Ubiquitous unique, beautiful art. Entire movies made by machines. Too many to even watch them all. Actor models that aren't real people owned by th…

Images generated have a watermark that can be used to filter.

The images I generate don't have a watermark. Same applies for a lot of other AI generated images. The watermark is the most trivial part to remove and is definitely not a defining feature of what makes an image AI generated.

Re: Stable Diffusion is a big deal

#55

Post by thewanderer1983 »

throwaway1851 wrote: post #8
LewisVerstappen wrote: post #3

Earlier quoted context omitted.

Why do people keep bringing up copyright anyway? It seems pretty clear that the images being generated by StableDiffusion are transformative so it's protected under Fair Use. Am I missing something here?

There are really significant, novel copyright issues implicated by these large generative models trained on other people’s IP. If you take a step back, you can see that there are different ways to frame what is happening. One frame is: “Defendant built an algorithm that memorized features of Plaintiff’s IP. Defendant’s algorithm recombines parts of those features in order to produce works in the same domain that comp…

Artist today use the exact same method of learning from other peoples artwork to generate new artwork and styles. These models are learning just like any artist learns and then producing new content.

Re: Stable Diffusion is a big deal

#56

Post by choeger »

> No-one expected creative AIs to come for the artist jobs first, but here we are!

Maybe that's because we never really thought about it. In hindsight, it's only logical. For an artistic rendering, correctness doesn't matter much nor does understanding the model. For flying a plane or driving a car or just transforming code from one language to another, it very much does.

Re: Stable Diffusion is a big deal

#57

Post by bowsamic »

ryandv wrote: post #22

Quoted post unavailable.

Pretty much any ethical discussion about race is suppressed on HN because this is a “Silicon Valley neo rationalist” site

Liam Kofi Bright outlines this with the Represser archetype in his new essay “White Psychodrama”

Re: Stable Diffusion is a big deal

#58

Post by aquafox »

Stabile diffusion is most of all a waste of resources! 20 billion parameters trained, using I dont know how many CPU hours, all for making better artistic images than what an amateur would do and on par with professionals. Art is nice, but not important for survival, contrary to smartly using the resources of the planet. Do you want to see a future way to live on or "the rest of Mona Lisa"?

Re: Stable Diffusion is a big deal

#59

Post by trention »

thewanderer1983 wrote: post #55
throwaway1851 wrote: post #8

Earlier quoted context omitted.

There are really significant, novel copyright issues implicated by these large generative models trained on other people’s IP. If you take a step back, you can see that there are different ways to frame what is happening. One frame is: “Defendant built an algorithm that memorized features of Plaintiff’s IP. Defendant’s algorithm recombines parts of those features in order to produce works in the same domain that comp…

Artist today use the exact same method of learning from other peoples artwork to generate new artwork and styles. These models are learning just like any artist learns and then producing new content.

Even if that's true (which it isn't), it doesn't matter. It's absolutely logically consistent to allow humans to do it while forbidding AI to do it.
Post reply on HN