@freya@social.highenergymagic.net @akana@tech.lgbt I'm very heavily anti current LLMs, and to be completely honest this is definitely a tough question. I can, of course, see how the accessibility improvements can be huge... but I can also see very strong arguments for why they still don't outweigh the harms of corporate LLMs. Accessibility doesn't have infinite value from a utilitarian standpoint, and oh boy do these machines of copyright violation make a very strong case for how an unethical enough tool can outweigh the benefits of accessibility.

Overall, I still find any of the modern models in use incredibly unethical, and I sincerely hope that they crash and burn for their horrific training data practices. The future of ethically trained models is bright if we just cut the techno fascists who are perfectly happy to be entirely unethical out.