News for hackers. Stuff that matters.
Re:Papers on GitHub Copilot, copyright law, and ownership fo... by visarga on Saturday June 25, 2022 @02:20PM (#31877476)
A zip file is not a good analogy to a language model as the zip can reproduce the original exactly with 100% recall, language models have very few episodes of verbatim reproduction and coverage is spotty. A zip file can only reproduce the original, but a language model can be asked to write custom, situated code. The language model still works if you block all the training data from appearing in its output.
Re:Papers on GitHub Copilot, copyright law, and ownership fo... by trention on Saturday June 25, 2022 @02:52PM (#31877709)
Potential output has nothing to do with the question whether the training data is "memorized" in the weights or not. To me, the answer to the latter question is "obviously yes".
Re:Papers on GitHub Copilot, copyright law, and ownership fo... by aaaaaaaaaaab on Saturday June 25, 2022 @03:44PM (#31878122)
Like I said, it's a lossy form of compression.