[satvikpendem]

Why would a derived image be under copyright?

[insanitybit]

Because that's legally the case? I don't understand the question. Using a lossy compression algorithm on an image does not remove its copyright protection.

[satvikpendem]

Taking in an image then creating something new based on the ideas of that image is legally permissible, if you mean that you think AI training produces only derivative works.

[insanitybit]

I am not sure what you're trying to say. Compressing an image is creating a derived work, it is still subject to the copyright of the original.

[satvikpendem]

What AI is doing is not analogous to compression so even if you think they are they same, it legally and technically speaking is not subject to the copyright of the original.

[insanitybit]

Are you missing the comment that I'd responded to?

> Information entropy. The amount of data an LLM ingests cannot be compressed to the size of the weights even at maximum theoretical compression.

[satvikpendem]

I wrote that comment. As I said, an LLM's weights are less than the size of theoretical compression, therefore what LLM training is doing is not compression at all.