3 ms·
Ok, so let's assume it's using a Visual Transformer architecture, since AFAICT, we don't know. What the model does is turn the picture into numbers in a clever
by RC_ITR 4y ago
Ok, so let's assume it's using a Visual Transformer architecture, since AFAICT, we don't know.
What the model does is turn the picture into numbers in a clever way (embedding).
Then, the model compares the relationships between all the embedded tokens (self-attention).
Then the model does the fancy linear algebra stuff we've come to know and love (feed forward neural network).
What you're basically getting at with your temperature comment is "rather than having it semi-randomly pick one of the top n options that the feed-forward network recommends for which pixel goes where, we can always have the model simply choose the top answer each time."
That is not "we can make it reproduce training images"
The point of embedding is to see multiple examples of similar images and generalize from them. The process in and of itself breaks the ability to re-create a photo, unless the training set is very skewed.
If that's the argument of legal cases, they are likely to fail or win on ignorance.
- blibble 4y ago> That is not "we can make it reproduce training images" didn't say that, all I said is it's a deterministic lossy compression function of the inputs (unless randomness is deliberately introduced) and as such it should be treated the same way as zlib or jpeg for copyright purposes
- RC_ITR 4y agoI think you're really stretching the definition of deterministic here. For example, switching training order of images has a known impact on outputs, but I don't see the equivalent in traditional compression. I agree with you that it is computer code that runs on circuits (and is therefore deterministic in a very loose sense), but it is fundamentally incorrect to compare it to traditional image compression.
- f_devd 4y ago> deterministic lossy compression function This would only be true if it didn't mix or 'forget' any of the training data for each prompt, and only then for the exact same inputs of the training data. That is a lot of qualifiers that aren't practically applicable. Slight side note: a deterministic lossy compression function should not always be treated the same as jpeg/zlib, as hash functions are technically also in that category.
- pertymcpert 4y agoThose are two orthogonal concepts. The temperature has nothing to do with whether it's a lossy compression function.
- Robotbeat 4y agoCopyright doesn't have anything to do with determinism. It existed in a pure analogue world before computers and if you had implemented your neural net in a fundamentally non-deterministic way (such as an analog neuromorphic neural net of some sort), it wouldn't make any difference one way or another. A human artist with no computer whatsoever and only their brain and basic writing (or painting) implements can break copyright just fine by writing out (or painting) copyrighted works from memory, then distributing them. It's the PRODUCTION side that is the copyright infringement, not the observing side.
- avisser 4y agoI have no idea why your replies seem to think you're being outrageous. I've seen the image from one of the AI services that is a very clear almost-reproduction of a Getty Images photo. I wouldn't want to be the defense's lawyer when that image gets shown to a jury. https://www.theverge.com/2023/2/6/23587393/ai-art-copyright-lawsuit-getty-images-stable-diffusion https://www.theverge.com/2023/2/6/23587393/ai-art-copyright-...