I don't use LLMs for anything meaningful (I have tried them, just to help understand how they work). I think the reason is something like this:
1. In would be really keen on a computer <-> brain interface. The possibilities are fascinating.
2. But only if it was:
- open source,
- understandable (not black-box), and
- not corporate-controlled.
So neuralink is a fuck no. I think the visceral discomfort with and repulsiveness of the idea of someone like Elon Musk having direct access to your brainstem should be obvious to anyone half-sane.
3. LLMs are not quite the same thing, but they are definitely in some kind of uncanny-valley of similarity. The link is linguistic, and not electrical, but that's just one level of abstraction away. And they fail all three criteria in point 2 (at least all of the commercial ones do, I don't have the energy or impetus to set up my own, I'd rather just do the learning myself).