I think that there are some people working on this, and a few groups that have claimed to do it, but I’m not aware of any that actually meet the description you gave. Can you cite a paper or give a link of some sort?

Y’all should really stop expecting people to buy into the analogy between human learning and machine learning i.e. “humans do it, so it’s okay if a computer does it too”. First of all there are vast differences between how humans learn and how machines “learn”, and second, it doesn’t matter anyway because there is lots of legal/moral precedent for not assigning the same rights to machines that are normally assigned to humans (for example, no intellectual property right has been granted to any synthetic media yet that I’m aware of).

That said, I agree that “the model contains a copy of the training data” is not a very good critique–a much stronger one would be to simply note all of the works with a Creative Commons “No Derivatives” license in the training data, since it is hard to argue that the model checkpoint isn’t derived from the training data.

Model sizes are larger than their training sets

Excuse me, what? You think Huggingface is hosting 100’s of checkpoints each of which are multiples of their training data, which is on the order of terabytes or petabytes in disk space? I don’t know if I agree with the compression argument, myself, but for other reasons–your retort is objectively false.

I’m getting really tired of saying this over and over on the Internet and getting either ignored or pounced on by pompous AI bros and boomers, but this “there isn’t enough free data” claim has never been tested. The experiments that have come close (look up the early Phi and Starcoder papers, or the CommonCanvas text-to-image model) suggested that the claim is false, by showing that a) models trained on small, well-curated datasets can match and outperform models trained on lazily curated large web scrapes, and b) models trained solely on permissively licensed data can perform on par with at least the earlier versions of models trained more lazily (e.g. StarCoder 1.5 performing on par with Code-Davinci). But yes, a social network or other organization that has access to a bunch of data that they own, or have licensed, could almost certainly fine-tune a base LLM trained solely on permissively licensed data to get a tremendously useful tool that would probably be safer and more helpful than ChatGPT for that organization’s specific business, at vastly lower risk of copyright claims or toxic generated content, for that matter.

The problem with your argument is that it is 100% possible to get ChatGPT to produce verbatim extracts of copyrighted works. This has been suppressed by OpenAI in a rather brute force kind of way, by prohibiting the prompts that have been found so far to do this (e.g. the infamous “poetry poetry poetry…” ad infinitum hack), but the possibility is still there, no matter how much they try to plaster over it. In fact there are some people, much smarter than me, who see technical similarities between compression technology and the process of training an LLM, calling it a “blurry JPEG of the Internet”… the point being, you wouldn’t allow distribution of a copyrighted book just because you compressed it in a ZIP file first.