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

371–380 of 488 posts

Re: Stable Diffusion is a big deal

#371

Post by earth_walker »

lelag wrote: post #342
pclmulqdq wrote: post #330

Earlier quoted context omitted.

If you use my copyrighted code to train your brain and reproduce exact copies of functions I wrote, I have a problem with that. Same for art. Exact copying is the copyright problem. If you use my copyrighted (public) code to train your brain, and then produce new functions that do new things, there is no problem. Replace "brain" with "AI" and you get my position on this (and a lot of others' positions). It seems to b…

Indeed. A difference I see between the GPT-like language models and those image generation models is that it feels like the language models actually hold full copies of a lot of sample of the training dataset (hence their ability to recite existing content), whereas the image generation models clearly do not: Stable Diffusion is 4Gb something, yet it can draw anything. That's the amazing part: the training dataset co…

I see what you mean about GPT returning copies. But is there really a difference if your model just happens to 'calculate' an exact copy instead?

For example, if you enter in something simple and high profile, what it returns looks pretty close to an existing work...

Try, for example, typing in "Banksy", "Brad Pitt" or "Starbucks".

So if you type in "Photo of a Coffee", "Impressionist painting of a fruit bowl" or "Blue canvas with one red line in the middle", how do you guarantee that the image you get back isn't actually a copy of someone's work?

Re: Stable Diffusion is a big deal

#372

Post by odessacubbage »

Workaccount2 wrote: post #336
peoplefromibiza wrote: post #326

Earlier quoted context omitted.

> not any more-so than I stole from him when I looked at that image I spoke with the author, the author disagrees with you. The reason he consider that stealing is because they took that image from the online shop where the disclaimer specifically states "2022 © Magnetic Press LLC by Arenathemes. All Rights Reserved" without even trying to contact the copyright holder. https://store.magnetic-press.com/products/viewpo…

I would venture to guess that the author doesn't have a full understanding of what the AI is doing or how it works. There are a gazillion artist who take clear inspiration from other artists, and there is no copywrite violation in doing so. The AI viewing his work isn't any different.

using work without permission is a violation copyright. imo the fact that the original has to be scrapped to create the dataset completely negates any argument about whether the final product is transformative b/c it cannot exist without the original data

Re: Stable Diffusion is a big deal

#373

Post by roywiggins »

shagie wrote: post #351
bsenftner wrote: post #271

Earlier quoted context omitted.

Exactly, this replaces general illustrative commercial art. This is great for illustrations at the head intro of fiction stories. However, fine artists that produce art where the object created is merely the totem for their actual art: intellectual conceptualization of an idea, often concerning the finality of life and the immature behaviors we playout within; that type of Fine Art is completely out of reach of moder…

I can see this replacing the clip and misappropriated art in slide presentations that no one is paying for currently with a $20/month service that lets you do "line art angry man in toga at computer" to put into your talk on Kuberentes. It might replace the "nice pic, do you have it at {size} so I can use it as a desktop image" comments on various social media sites. I don't see it replacing an actual photograph as a…

> because they are subtly wrong in some ways

Well, they are now. Give it a year and we'll see.

Re: Stable Diffusion is a big deal

#374

Post by lurquer »

earth_walker wrote: post #371
lelag wrote: post #342

Earlier quoted context omitted.

Indeed. A difference I see between the GPT-like language models and those image generation models is that it feels like the language models actually hold full copies of a lot of sample of the training dataset (hence their ability to recite existing content), whereas the image generation models clearly do not: Stable Diffusion is 4Gb something, yet it can draw anything. That's the amazing part: the training dataset co…

I see what you mean about GPT returning copies. But is there really a difference if your model just happens to 'calculate' an exact copy instead? For example, if you enter in something simple and high profile, what it returns looks pretty close to an existing work... Try, for example, typing in "Banksy", "Brad Pitt" or "Starbucks". So if you type in "Photo of a Coffee", "Impressionist painting of a fruit bowl" or "Bl…

Copyright - at least for images - is going to become moot shortly.

Copyright exists to allow an artist time to reap the benefits of his labor. Without this time, they say, no rational person would invest his labor into making art. Dubious… but, whatever. The point is, if you remove the ‘labor’ from the mix, there’s no need for copyright.

If I can produce spectacular images with zero individual labor, there will be little reason for me to copy someone else’s work.

Re: Stable Diffusion is a big deal

#375

Post by lurquer »

The underlying data, I gather, must associate an image with a description. There aren’t millions of well-written descriptions of images. Yet, they’ve accomplished this with, I think, with merely the sparse text that comprises the title of the scraped JPEG or perhaps the surrounding text on a website.

However, shortly, there will be millions of extremely detailed descriptions of new images… that is, a person puts in a detailed prompt, and if you can capture how pleased the user is with the result, you could then add that new picture — along with the user’s prompt — to your training data. Eventually, you will have millions of well-described images, which will make the system even more amazing.

Re: Stable Diffusion is a big deal

#376

Post by earth_walker »

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…

I think there will likely be better ways to bias the training of specific instances - Something like a 'training library' with biased training data that you can plug in for your use case.

For example let's say you're a well known designer with a distinct style, and you train your own instance on your lifetime's body of work. Now you can generate whatever you want in 'your style' (just like you can now ask for a painting in Dali's style).

Now you've turned your style and design process into a factory - for every new client you can create whatever they want, in your style, with multiple examples, in a button press. Perhaps you can even sell that 'training library' to other designers?

Re: Stable Diffusion is a big deal

#377

Post by ed »

Tomis02 wrote: post #349
habosa wrote: post #121

These new models completely changed my mind about how much impact AI will have in my lifetime. They are the most impressive software achievements in decades and anyone who has a “meh” reaction will absolutely end up looking silly. I’m not optimistic that their impact will be positive though.

> They are the most impressive software achievements in decades and anyone who has a “meh” reaction will absolutely end up looking silly. With the risk of looking silly, I declare "meh" once more, just as I "meh"-eh when GPT-3 came out. The so-called "AI" is not fundamentally different from the AI of the 80s, it's just that now we have much better hardware. The main problem of the past AI winters still remains - the…

incremental improvements are easy to dismiss but it can be hard to tell when some critical mass of utility is reached. The first cars were electric (and steam-powered) but it took 100 years to supplant ICE (steam is next /s). the components for drone technology aren't new but control systems and batteries needed time to improve, VR has incrementally improved since the 80's, even the internet was around for 20 years before incremental improvement brought us the web...

Re: Stable Diffusion is a big deal

#378

Post by winddude »

_gabe_ wrote: post #307

Is it just me, or do the comments in this thread seem to be the exact opposite of the sentiment in the comments on similar Github Copilot threads? I just find it a bit ironic that programmers are irate about Github Copilot using their copyrighted material to train. However, if it's an ML model training off of copyrighted artists material, clearly its a transformative work. I just find the opposing sentiments for thes…

Copilot wasn't released opensource.

Re: Stable Diffusion is a big deal

#379

Post by ssizn »

ManuelKiessling wrote: post #318
ssizn wrote: post #283

Earlier quoted context omitted.

I tried to find where the implementation in your code is but there are like a thousand files spread over a hundred folders.

You... are trying to troll me, right? % find . -type f ./migrations/Version20220723092633.php ./composer.lock ./.env.local ./bin/connect-to-db.sh ./bin/_init.sh ./bin/deploy.sh ./bin/visualize.sh ./bin/console ./config/services.yaml ./config/bundles.php ./config/packages/messenger.yaml ./config/packages/framework.yaml ./config/packages/doctrine.yaml ./config/packages/translation.yaml ./config/packages/cache.yaml ./co…

You conveniently omitted all folders with gitignore files.

Re: Stable Diffusion is a big deal

#380

Post by SyzygistSix »

systemvoltage wrote: post #359
SyzygistSix wrote: post #261

Earlier quoted context omitted.

No, apparently we just massively underestimate how important studying the humanities is. This is illustration, not art. I'm flabbergasted how many people are so quick to confuse or conflate the two.

For the record, my significant other is an artist and my entire life has been surrounded by creative people. I’m questioning human creativity, say, 50 years from now, and how much machine creativity we currently underestimate. Sure these artworks lack imagination and fidelity today. Appreciating art and humanities has nothing to do with projecting AI capabilities in the future, I’m flabbergasted how many people are s…

50 years from now, I have little doubt that will be the case, barring serious civilizational decline. But that isn't the sentiment on this page.

Ditto for me regarding being surrounded by creative people. This isn't going to lose them any business any time soon, not even the younger ones five years from now.

Post reply on HN