Esme Povirk @madewokherd@computerfairi.es

I am beginning to suspect that the labor exploitation in how LLMs are trained (tuned?) has a detrimental impact on the quality of responses. Basically, LLMs sound like someone who is being exploited, which comforts and discomforts entirely the wrong people.

I have set up a local LLM with a web UI on my laptop. It was pretty easy (for that I credit Arch/AUR). It barely runs but my laptop doesn't have a fan or discrete GPU so it's kind of amazing that it works at all.

(For the few things I use this for, I don't want to be there when OpenAI starts rent-seeking, and I'd rather not continue to inflate their numbers now.)

If we take the "it'll be too hard to tell who you're talking about" argument against they/them personal pronouns to its logical conclusion, we arrive at "everyone should have their own unique pronoun set". Which, has a certain appeal. Or maybe I just want my own unique pronoun set.