I’ve gotten a fair amount of feedback on my post about LLMs and copyright, most of it expected. But I was surprised to see how many people wrote to me to inform me that copyright violation is not stealing under natural law, that stealing is an inappropriate word to use. Copyright is a societal convention only, I have been told, an invention of the government with no moral force.

I was surprised by this argument, but I shouldn’t have been, as it is in line with the δόγματα of Richard Stallman and the Free Software Foundation, as well as off-shoots of that copyright protest strain of thought including the Pirate Party in its many forms, and Pirate Bay, etc.

I will simply say that, while I am sympathetic to avoiding the word “steal” for non-commercial piracy such as torrenting, I stand by the notion that making huge profits through illegal derivatives of creative works without giving a cut to the original artists and writers is, in fact, stealing. I disagree here with the FSF on a philosophical level. The moral force of the concept of “stealing” is indeed defined partially by natural law and human rights, but it is also defined by societal convention and, by extension, positive law.

It was also suggested that I release the contents of this blog under the GNU Free Documentation License. I am declining to do so. I note that not even the GNU licenses page would have me do so for the majority of my articles, which are works of opinion more so than works of education. Even for the ones that are works of education, I see them as creative works more so than works of documentation. I would consider a much more restrictive CC license, but am not going to do so at this time.

To be abundantly clear, I would not want even a fully “free”/“libre”/“open source” large language model trained on my articles. I am not convinced that that would be substantially more good for society than the commercial ones would.