@nprofile1q... I've sometimes thought about things I could do with eg. local LLMs to try to make a user interface that isn't just yet another deceptive chatbot. I can think of several interesting ideas that there's a part of me that would like to explore.
But...
I can't really get over the stink of LLM, you know? Even if a local model can't surveil you and doesn't require you to pay money to the most awful people on the planet, they're *still* built from DDoS'ing the human internet, data taken without consent, and trained in data centres with a gruesome environmental cost.
I suppose you can argue that once the lab publishes the model, all that damage is already done. It just still feels wrong, you know? Like buying goods from somewhere with a terrible human rights record.