Towards Understanding Sycophancy in Language Models

by fzliuon 6/16/2025, 4:19 AMwith 2 comments

by bitwizeon 6/18/2025, 1:46 AM

Putting Genuine People Personalities into LLMs turned out to be a bad idea.

Seriously, though, as I said years ago it's much like the Asimov story "Liar!" The robot is trying to please you by telling you what you want to hear.

by gnabgibon 6/16/2025, 4:27 AM

(2023) Discussion at the time (57 points, 72 comments) https://news.ycombinator.com/item?id=38016013