111–120 of 148 posts
Earlier quoted context omitted.
Midjourney uses SD under the hood (you can see in their license), but they augnment the model in various ways.
The results in midjourney are significantly better than SD. I find it much easier to get to a good result in MJ and I've been trying to understand why. Anymore insight you could share?
So, and this is an ELI5 kind of question I suppose. There must be something going on like "processing a kazillion images" and I'm trying to wrap my head around how (or what part of) that work is "offloaded" to your home computer/graphics card? I just can't seem to make sense of how you can do it at home if you're not somehow in direct contact with "all the data?" e.g. must you be connected to the internet, or "stable…
Earlier quoted context omitted.
Is there a way to run this in the cloud? On Google Colab or elsewhere?
Colab: https://colab.research.google.com/github/WASasquatch/StableD... To run it elsewhere in the cloud, grab a GPU (spot) instance and SSH in.
I’m waiting for someone to wrap this up into a desktop app that I can install and run on my Mac.
I've been looking into this for the last 2 days. Unless you're running an M1 Mac or newer, you're SOL. Stable Diffusion is built on PyTorch. PyTorch mainly has been designed to work with Nvidia cards. However PyTorch added support for something called RocM like a year ago that adds compatibility with newer AMD cards. Unfortunately RocM doesn't support slightly older AMD cards in conjunction with intel processors. So…
Unless you want to train the model, Lambda Labs is somewhat cheap:
Earlier quoted context omitted.
>as they are asking if the AI itself can hold the copyright on the output. Who is? The original question I replied to: >If someone runs the model on their own hardware, do they "own" the images generated? This seems to be straightforward - Thaler tried to receive the copyright for the artwork generated by his Creativity Machine. He was denied, because the copyright office does not believe that a neural network genera…
> > as they are asking if the AI itself can hold the copyright on the output. > Who is? Thaler is. I’ve only read the intro sections of the documents you linked to, so I may have missed something more fundamental later, but the key points seem to be: > The author of the Work was identified as the “Creativity Machine,” ... the Work “was autonomously created by a computer algorithm running on a machine” and: > Thaler m…
>In that case, a human is providing a prompt to the model (providing creative input), and asking if they themselves count as the author (a human rather than a neural net), so it seems like a significantly different case.
I don't know that I specifically agree with this, but this is probably due to me having read additional articles on similar filings, including one where someone took a photograph, applied a style transfer AI to it, and then tried to copyright the resulting image, and was denied, because the copyright office found that there was not evidence that the work was a product of human authorship.
Andres Guadamuz (a lawyer specializing in IP law, senior lecturer at Sussex university, and a proponent of AI generated work being copyrightable) discusses a lot of this in https://www.technollama.co.uk/dall%c2%b7e-goes-commercial-bu... - but the most relevant part to this discussion is "For the most part, the legal consensus appears to be that the images do not have any copyright whatsoever, and that they’re all in the public domain."
The user experience for DALL-E, StableDiffusion, Midjourney, etc. are all essentially the same - craft a prompt, fine-tune it, get artwork out, so his discussion should be broadly applicable to all of these similar tools.
Earlier quoted context omitted.
This is exactly it. It’s pretty remarkable that it was trained on over 100 terabytes of images and yet the model has been distilled down to only 4gb.
Then maybe we should remind about this 25,000:1 ratio when an artist complains about his copyrights being abused. The model doesn't have space to actually copy his works inside, it can only memorise the equivalent of a thumbnail from each input. A very small thumbnail, scaled down 150:1 per width and height (square root of 25000). That's like a grain of rice on the screen.
When programming, it will often take a long time and a lot of code to get to a few final lines that do what you want. You cannot say the final result is a "thumbnail" of all previous efforts. Rather, it is the apotheosis of it.
Some artists spend decades developing a style that looks like a kid could do it as well. Still, there is something unique in there, that a trained eye will recognize. Converting that particular style to a formula and making that freely available is at least somewhat morally ambiguous.
How does copyright work with output images? If someone runs the model on their own hardware, do they "own" the images generated? If 2 people generate the same image using the same prompt/seed, who "owns" the image?
Is it any different than Photoshop content aware fill? Or using a camera? Nobody would ever think about Adobe or Nikon having copyright claims over your pictures. For me it's just a tool, the artistic part is providing a good description/base image, refining and choosing the best output. Anyway, I'm not a lawyer and we probably live in different countries, so it'll be interesting to wait for the first lawsuit.
But it is illegal to share pictures of Eiffel Tower, for example.
People do it, but they shouldn't.
If I put a picture of Eiffel Tower at night in a book or any other kind of commercial product, I have to pay to use it. Doesn't matter that it's there for my eyes to see it.
The question is: are the images generated by an hyper accelerated learning machine using copyrighted material without the author's consent legal?
I think they shouldn't be and the data included in the training should be free or licensed.
Earlier quoted context omitted.
Then maybe we should remind about this 25,000:1 ratio when an artist complains about his copyrights being abused. The model doesn't have space to actually copy his works inside, it can only memorise the equivalent of a thumbnail from each input. A very small thumbnail, scaled down 150:1 per width and height (square root of 25000). That's like a grain of rice on the screen.
That's not how it works though. Instead of applying arbitrary content detail reduction, the model is an attempt to distill the core of what makes a particular artist (or phrase, face, object etc) unique. When programming, it will often take a long time and a lot of code to get to a few final lines that do what you want. You cannot say the final result is a "thumbnail" of all previous efforts. Rather, it is the apothe…
Earlier quoted context omitted.
My understanding is that images will not be bit identical due to GPU physics and decimal precision. Images from the same seed may be for all practical intents and purposes indistinguishable - but there are some flipped bits involved.
That's not my understanding. The same seed value to the device's random number generator should results in the exact same outputs - there's a bug being chased down in the MPS (MacOS) backend where the fixed random seed doesn't output the same image on different computers.
Earlier quoted context omitted.
> As it’s deterministic (given the exact same request parameters, random seed included, you get the exact same image) it’s a form of compression (or at least encoding decoding) too: I could send you the parameters for 1 million images that you would be able to recreate on your side, just as a relatively small text file. For any input image? Or do you mean an image generated by the model?
You could input an image and get it to recreate it as best as possible and then output a seed. That would be interesting!