Hacker News Reader

Comment by bigyabai | Hacker News Reader

https://en.wikipedia.org/wiki/Sycophancy_(artificial_intelli...

To be fair to the sycophancy tendencies, this was an open conjecture that held for 85 years, and not for lack of trying. So, maybe a bit warranted here? :)

I told GPT5.6 I came up with this walking down the street. "Jesus Christ" was its response.

Lol, you must have some custom instructions on that beast, lol

4.8 is much funnier: you have to convince it the counterexample is real, it keeps coming up with goofy excuses. As you talk it down, it keeps talking about its "discomfort" and "flinching".

That’s the worst thing about sycophancy. You can never tell when it’s warranted or not. I imagine that many AI chats have had real gold in them and yet we’ll never know.

The same was obviously true historically. People have had many revolutionary thoughts or even writings that nobody noticed. I think in this case it just feels worse because the AI could be the perfect tool for discerning the diamonds in the rough.

The interesting thing about using Claude Fable 5 is it's nearly as irritatingly sycophantic as past Claudes while genuinely being smarter than the previous models. So you get a kind of yo-yoing of it glazing you as a creative genius and disappointedly revealing to you that your ideas are bad and dumb.

That is easily fixable with the "Instructions for Claude" in your settings. Just put:

"be critical, but correct. I don't want affirmation, i want to actually accomplish things."

I don't get any sycophancy from it. The conversations are actually quite good, and it'll push back if it thinks I'm wrong.

gotta try with "think in russian"

The I do that it starts acting like a jerk.

I mean if someone came to you with a value of n that disproved collatz, wouldn't you go crazy as well?

Reply on news.ycombinator.com