6 ms·

Leaving aside what I think IP law ought to look like, LLMs literally train on (and reproduce!) the fixed, copyrighted expression. They aren't producing a new wo

Leaving aside what I think IP law ought to look like, LLMs literally train on (and reproduce!) the fixed, copyrighted expression. They aren't producing a new work based on abstract knowledge of algorithms that happens to share a particular expression. It's not at all obvious to me why the training set copyright wouldn't nominally flow through the model, even if that seems impossible to actually implement in practice.

visarga 4y ago
Why is everyone fixating on 1% of the outputs. 99% are sufficiently distinct. It replicates code as an unintended side effect, but there are mitigations.
- AlotOfReading 4y ago
  If you take all the words from Harry Potter and shuffle the words together in a random order, the output will be distinct but the work as a whole is probably still derivative. If you do that for 500 book series, you still haven't generated a new work, you've simply created a legal morass of derivative copyrights. ChatGPT is just a really impressive word-blender. All of those "sufficiently distinct" snippets are in a meaningful sense recombinations of other copyrighted snippets it was trained on. When you're in a very sparse part of the space or have a single work that's present many times in training like fast inverse square root, the model may even replicate recognizable chunks of the training set. It could be that this is sufficiently transformative that the output is not derivative, but as I said it's not obvious to me.
  - andybak 4y ago
    > If you take all the words from Harry Potter and shuffle the words together in a random order, the output will be distinct but the work as a whole is probably still derivative. If you do that for 500 book series, you still haven't generated a new work, you've simply created a legal morass of derivative copyrights. This might be the worst metaphor I've stumbled across in this whole debate. You're surely making an argument here for the opposite team.

visarga 4y ago

Why is everyone fixating on 1% of the outputs. 99% are sufficiently distinct. It replicates code as an unintended side effect, but there are mitigations.

AlotOfReading 4y ago
If you take all the words from Harry Potter and shuffle the words together in a random order, the output will be distinct but the work as a whole is probably still derivative. If you do that for 500 book series, you still haven't generated a new work, you've simply created a legal morass of derivative copyrights. ChatGPT is just a really impressive word-blender. All of those "sufficiently distinct" snippets are in a meaningful sense recombinations of other copyrighted snippets it was trained on. When you're in a very sparse part of the space or have a single work that's present many times in training like fast inverse square root, the model may even replicate recognizable chunks of the training set. It could be that this is sufficiently transformative that the output is not derivative, but as I said it's not obvious to me.
- andybak 4y ago
  > If you take all the words from Harry Potter and shuffle the words together in a random order, the output will be distinct but the work as a whole is probably still derivative. If you do that for 500 book series, you still haven't generated a new work, you've simply created a legal morass of derivative copyrights. This might be the worst metaphor I've stumbled across in this whole debate. You're surely making an argument here for the opposite team.