Silence Dogood
A. P. HowellSFWA's statement on AI/ML:
https://www.sfwa.org/2023/06/13/current-statement-on-ai-ml-use/
My two cents (as a member and volunteer but not part of the Board):
I think this is a good start. It lays out some of the complexities of the issue (e.g. terminology, lack of transparency) that make it challenging to make a statement that is definitive and pithy. It also approaches the issue from the perspective appropriate for the organization (i.e. how authorship and income are impacted).
Since this is only a start, I wish this particular statement had come out earlier this year, when generated content became more prevalent. I wish the terms of the discussion had been set then, so that now we could be a little farther along. I also wish the there had been a nod to the current legal landscape (e.g. decisions that machine-generated content is not copyrightable) but IANAL and I assume lawyers were involved in the drafting of the document.
Since this is only a start, I'm also eager to see future statements. The sort of things I'd like to see addressed include:
1. A discussion of terminology, possibly including a glossary for how SFWA defines/categorizes different types of technology.
2. Discussion of how AI/LLM/etc. will impact membership eligibility. (On the birdsite, Neil Clarke raised the point that, since membership is now earnings-based, there are no longer "qualifying markets.")
1. A discussion of terminology, possibly including a glossary for how SFWA defines/categorizes different types of technology.
2. Discussion of how AI/LLM/etc. will impact membership eligibility. (On the birdsite, Neil Clarke raised the point that, since membership is now earnings-based, there are no longer "qualifying markets.")
3. How AI/LLM/etc. issues will be addressed by SFWA resources and projects (e.g. Writer Beware, Griefcom).
This statement provides hints of how policies and future statements may develop, and I'm encouraged by the mention of an Emerging Technology Committee.
mirabilos@aphowell this is so much better than the shit stance of @[email protected] which reminds me to cancel my recurring donation to them.
@mirabilos ...and Wayback Machine for the win, to find the interview in the excised bit.
On the one hand, I can understand that with "transformative" in the name they might want to step lightly. On the other, IIRC there's a whole lot of "our stuff falls under fair use because it's transformative," which is also an argument for it being protected by copyright.
@mirabilos The assertion that there are training sets based only on public domain works may be true, but the movement (sets began as public domain, then expanded) is at odds with my understanding of how some of these companies work, and it's certainly not how earlier mass digitization projects have worked. And the idea that fanfic will fix the racism/misogyny/etc. baked into these products is frankly laughable. (IANAL and IANA ML expert, but I was also not born yesterday.)