I’m brand new to Nix, and I was just checking out the release schedule (NixOS 24.05 — Release Schedule · Issue #303285 · NixOS/nixpkgs · GitHub) to figure out if I should hold off for the new release (I prefer sticking to stable branches) or not. While browsing, I stumbled upon a comment regarding Jon’s ban, which piqued my interest to learn more about how the project operates. I’m keen on diving into NixOS, so understanding why Jon got banned, whether it was justified, and if I want to be part ...
Is this nonsense spreading? Stop trying to bring back forum signatures mixed with ‘I do not give Facebook permission, repost this before midnight.’ Legally, practically, and morally, it is incorrect.
If you could reserve those rights, they would be so by default.
Properly mangling information through a deep neural network is about as transformative as fair use gets. Doing it wrong and just storing data is a failure case. (And it suggests a lot of wasted effort, where that data could be referenced instead of trained. Ideally we could show the LLM a new textbook and have it explain by reading, rather than stirring the textbook into the zillion-dollar back-end process that created it.)
The loudest critics may not even know what they want. Guys: you expect the robot to answer specific questions about the US constitution and the characters in Harry Potter, while incapable of quoting from exactly one of them? Like it should mumble its way through paraphrasing “Yer a wizard… Barry,” but if it gets one word wrong in the second amendment then it’s useless. We finally created a machine that speaks English and people are mad it doesn’t immediately grasp copyright law. When most people don’t.
Is this nonsense spreading? Stop trying to bring back forum signatures mixed with ‘I do not give Facebook permission, repost this before midnight.’ Legally, practically, and morally, it is incorrect.
If you could reserve those rights, they would be so by default.
OpenAI et al. will just put “No copyright infringment intended” on everything and it will all be fine.
Honestly, that’d be equally valid.
Properly mangling information through a deep neural network is about as transformative as fair use gets. Doing it wrong and just storing data is a failure case. (And it suggests a lot of wasted effort, where that data could be referenced instead of trained. Ideally we could show the LLM a new textbook and have it explain by reading, rather than stirring the textbook into the zillion-dollar back-end process that created it.)
The loudest critics may not even know what they want. Guys: you expect the robot to answer specific questions about the US constitution and the characters in Harry Potter, while incapable of quoting from exactly one of them? Like it should mumble its way through paraphrasing “Yer a wizard… Barry,” but if it gets one word wrong in the second amendment then it’s useless. We finally created a machine that speaks English and people are mad it doesn’t immediately grasp copyright law. When most people don’t.