/r/LocalLLaMA/ is such an interesting mix of academia/researchers (it was called out in a recent paper, regarding context length, IIRC) and odd anarcho-futurist weirdos. And I kind of love it for that.
Ha Stable Diffusion is very similar I notice. People understanding and thriving in new techniques and advanced workflows, mixed in with people who want to generate boobies in new ways.
I found they can be very unappreciative of their data sources for training. Any mention that scraping art from art sites like DA without their consent was unethical is met with arguments about how its exactly the same as a human looking at art and replicating the style.
They refuse to even acknowledge why those people might be annoyed their work was used without their consent or compensation to put them out of work.