[stopthe]

Unfortunately, the choice of license likely won't matter in the nearest future (if not already so). If a tech giant wants you open-source library, they will just point their agent to it and ask "to rewrite in the style of War and Peace". And more unscrupulous players won't even bother with a rewrite, as we've seen recently in the case of Cheatingdaddy/Pickle.

[roenxi]

The flip side is that if they can technically pull that off then the cost of writing the library has dropped so low that an OSS maintainer probably wouldn't have to work too hard to write it anyway.

[pimterry]

Being able to rewrite existing working code sufficient to copyright-launder it isn't the same as being able to write it from scratch, unfortunately, especially since LLMs seem to be allowed to ignore quite a bit of copyright law with complete impunity.

Imo it's totally plausible that something will be expensive & time consuming to create, even with LLMs, but still easy to fork outside current licensing restrictions with LLMs.

[layer8]

Rewriting it with a guarantee of not introducing any errors is still beyond current LLM capabilities, and there might be a certain correlation between that capability and the capability of writing it from scratch.

[rcxdude]

>If a tech giant wants you open-source library, they will just point their agent to it and ask "to rewrite in the style of War and Peace"

Is there any evidence of this happening? And any legal theory behind how it might have the intended effect? Training being fair use does not make AI a magical copyright-removal box.

[krapp]

AI is already "too big to fail" in every conceivable way. If there is a conflict between AI and the law, then the law will inevitably be rewritten, reinterpreted or repealed in the way most beneficial to the service of AI.

[Dylan16807]

I expect a lot of leniency when it's something actually important to making AI work.

Training on most of the data in the world is important. Rewriting a document you don't own is not important. Or depending on what level the objection is made at, stealing some random library is not important.

[gtirloni]

what has been your experience with asking AI to create a complex project from beginning to end?

[stopthe]

Exactly zero, if answering your question as stated. But that is not the point. LLMs struggle with things that require tacit knowledge or are not found in the training set (i.e. haven't been done before). They excel at tasks that are described in the most verbose and explicit way, to the point that it looks humiliating from human POV. And what is more explicit than an existing codebase, especially with tests, comments and documentation lovingly put there by the proud author (remember, those github stars).

[gtirloni]

Thanks, that does answer my question.