LLMs are not compression algorithms. From an information theory perspective, that's impossible given their size.
Thus, a distinction needs to be made between viewing material to _learn_ and viewing material to _verbatim repeat_.
It's not illegal to read the New York Times and then start giving paid advice based on what you learned, as long as you don't repeat the text verbatim.
It’s irrelevant if it’s legal today or not. This is new technology and may be new precedent.
I can see how LLMs can't be lossless compression algorithms, but why not lossy?
... but you have to pay to read the NYT. You paid for the information. Guess who didn't.
Citation needed. Of course they are compression algorithms