seems this specific case, the dude was a little nuts, but did the chatbot somehow encourage or lead him to the breaks he had. my experiance is yes the gemini is too friendly, too positive, but i see it. it seems like a suckup. i tried to tune it by asking it to be more critical, but then it seemed to be an asshole. also tried the so-called “potato” profile, but again, too agressive. i like the positive feedback, but just want to tone it down a bit. regardless, it just seems so obvioius to mae that i ignore it (or, at least i THINK i am ignoreing it….); also, i’ve not had long conversations. the long ones are all in the pursuit ofd installing some fickiung pc of software that doesn’t quite work, many hours of going round and rond to solave a niggling little problem in a container in a docker on a linux serfver.
fta:
Marple adds that users will spiral after their long conversations with a chatbot, whatever model it may be, because, he thinks, companies can’t afford to do it differently. He argues sycophancy is a feature, not a bug. … He adds that, unlike human conversations, which feature pushback and different perspectives tugging at each other, a user doesn’t receive any pushback during their conversations with chatbots: “The design of the product is pushing you away from reality. It’s pushing you away from other people,” he said. “The friction with other people is what keeps us grounded.”
