> as regards posts online, left out the part where you fret (or do not know enough to fret) about the fact that you not knowing if it is a bot or not is part of the problem, or how it could affect you if you reply or not. many times, so many bots and drones have been unleashed looking for a juicy person to take down or troll that you find it difficult to interact online under your own identity, then find it difficult to show you have an online presence when you need to without worrying that that identity will get taken down a notch or three if someone who knows your online handle interacts with one that then goes after their contacts, and so on.

Yep, I just went from the top of my head and I left a lot of other negative experiences out. I think a lot of people are moving to burner accounts or otherwise choosing to not interact under their real identities - it's just too easy to spin up automated harassment now.

> the part about medical "hallucinations" is especially spot on. my last attempt at an ai-enabled zoom session was so wrong it got almost every identifying thing about me wrong, read my blood results wrong even though the outside of normal range color guidance was there, then near as I can tell the person after me was (redacted) and showed me as (redacted) (I have had at least five blood panels this year; i am not (redacted)).

Another fun one: a friend of mine said he "was driving a car to Sedona" during a phone call describing a neck injury several months ago. His last two follow up appointments have both had a note to "follow up on the carcinoma". Now that it's in the context, it seems impossible to get LLMs to ignore it, because "carcinoma" is an important word. A human would not have this problem.