With Stable Diffusion it's storing the style, but can't reproduce any single input image-- there aren't enough bits [0]. (except by luck, but that's really true for any storage).
Rank all generated images by similarity to the training data (etc.) and you can see what's stored.
The Shannon-Hartley theorem isnt relevant. A 4GB zip of 100TB text data can exactly reproduce the initial 100TB for some distributions of that initial dataset.
But if you reproduced an image whose style matches another copyrighted image (e.g., blah in the style of starry night), then how does that new image (which didn't exist before) violate existing copyright? You cannot copyright a style.
The NN containing information which _could_ be used to reconstruct an exact image doesn't itself constitute copyright violation - because the right to use information for training NN is not an exclusive right that the original holder of the training set has.
And if you can come up with a model that can reproduce images exactly without first getting trained on them, it wouldn't be a derived work, either.
but why is the set of numbers in a matrix considered derived?
I can trivially derive the number pi.
My original point is that just because something contains the information does not imply that it violates copyright.
You're correct that merely containing the information would not violate copyright - it's all about how that information was produced.
("If a human artist...")
A NN has not learnt to paint: it doesn't coordinate its sensory-motor system with its environment through play, it hasn't developed any taste, it does not discern the aesthetic good from the bad, it has no judgement, and so on ad infinitum.
A NN is just a kNN with an extra compression step. The way all gradient based "learners" work is to compute distances to pregiven training data. In the case of kNN that data is used exactly, in NN its compressed.
There is no intelligence here, there is no learning: it's a trick. It turns out that interpolating a point between prior examples can often look novel and often fool a human observer.
This is largely due to how incredibly tolerant to flaws we are in the cases where NNs are used to perform this trick. We go to great lengths to impart intention, fix communicative flaws, etc. and this is exploited by "AI" to make simple crap seem great by having the observer fill-in the details, perceptually. I see it as a kind of proto-schizophrenia that all people have which usually works if we're dealing with a human, but on everything else produces religions.
In any case, a NN is just a case of a kNN -- which is capable of fooling people exactly the same way, and clearly violates copyright and is a case of theft of intellectual work to make a product you can sell. Adding compression seems irrelevant.
Stable Diffusion is certainly capable of differentiating good from bad. That's why you can tell it to draw good or draw bad.
Not that this point is relevant to my comment. "Play", "taste" and "judgment" can be just as deterministic as a sequence of large matrix operations interspersed with nonlinear layers.
Interpolation is forced in the case of NNs, it's a training condition.
And the "kNN interpretation" isnt an interpration, kNNs define what "ideal learning" is in the case of statistical learning, and hence show, it doesnt count as actual learning.
In actual learning we're not interested in whether you can solve prespecified problems but how well you cope when you can't. This is, by definition, not a problem which can be formulated in statistical learning terms and the particular "learning" algorithm here is irrelevant.
In other words accuracy isnt a test of learning. Accuracy is a "non-modal condition" in being fit to a history that actually took place. Learning "in the usual sense" is strictly a modal, "what if" phenomenon, and is assessed by the quality of failure under adverse conditions, not of success.
If one gave any AI/ML system in existence adverse conditions, posed relevant "what ifs" and observed the results, they'd be exposed as the catastrophe they are. None survive any even basic test of "coping well" in these cases.
This is why all breathless AI public relations, ie., academic papers published in the last decade, do not perform any such tests.
you can create compressed copy of a file containing 100TB of the letter "A" in much less than 4GB
there could be enough bits in there to reproduce some of the inputs.
I wager 100 TB => 4GB is different from JPEG compression and more similar to what happens in our brains. "Neural compression" so to speak
Good thing this is already the case! https://en.wikipedia.org/wiki/Berne_Convention
> The Berne Convention formally mandated several aspects of modern copyright law; it introduced the concept that a copyright exists the moment a work is "fixed", rather than requiring registration. It also enforces a requirement that countries recognize copyrights held by the citizens of all other parties to the convention.
I was making the point that stable diffusion != JPEG compression
This is what we have courts and legislation for. I expect there's existing legislation here about what constitutes a different work versus an exact copy but it may need some updates for AI.
> the original images themselves aren’t stored in the Stable Diffusion model, with over 100 terabytes of images used to create a tiny 4 GB model
Is jpeg compression transformative then? Should a compressed image of something not be copyrightable because “it doesn’t store” the “real image”? How about compressed video? Where do we draw the line?