61–70 of 117 posts
> This suggests that people are drawn to AI that unquestioningly validate, even as that validation risks eroding their judgment The obvious counterpoint is the large number of complaints about OpenAI's excessively sycophantic models back in April 2026, which led to them rolling back to a previous and less sycophantic model. I think there are those who are susceptible to model sycophancy, but it's definitely not most…
This is true, but likely happening in talk therapy too with overly sycyphantic therapists
Sure, but there is a marked difference between a therapist and an LLM Chatbot: you can't talk to your therapist for +4h a day, at any time convenient to yourself. It's also not a guarantee you're going to hit upon a therapist that is like that, versus a product that is very literally intended to maximise your usage of it.
But that is only ~half of job. The other half consists of gently reminding one that not all is lost yet.
There is no LLM that can do this and I think there never will be.
Earlier quoted context omitted.
> if you're not seeing dramatic change, I suspect you're not looking. Ironically, this is the inverse of the very thing you complained about people doing to you.
Except I'm providing evidence.
That's a pretty far cry from the "evidence" you claim to have dropped on multiple comments.
Earlier quoted context omitted.
> Upvotes often do the same thing, especially in siloed communities. Yeah. I've gotten caught in this pattern here on HN. Post something, there's some disagreement but then I see the quiet upvotes trickling in, sometimes by the dozen. That sort of signal slowly convinced me I was right about plenty of stuff. I used AI to datamine all of my HN comments and find those patterns. Then I started asking the AI to steelman…
How would you say your views have changed since then?
AI measurably lowered the number of absolute claims I've been making. Tended to generalize and radicalize too much which would invite people to respond to the tone and the extremism instead of disproving the underlying idea. That sort of reply would actually reinforce my beliefs: if I was wrong, people would simply refute it. AI attacked my points with counterexamples until I was able to refine them into a core that's actually true.
These little distilled kernels of truth were the most valuable results. I found that many of my arguments actually have an underlying idea that holds up under scrutiny, but nobody cares because I try to prove it via extreme arguments. One such idea is anti-impossibility: laws and rules are all fine, but making it structurally impossible for people to disobey them is unacceptable. I used to try to argue that idea by defending running red lights which persuaded no one. Now I've identified the real idea and can make a much better argument for it.
> This suggests that people are drawn to AI that unquestioningly validate, even as that validation risks eroding their judgment I believe this is a larger problem than just AI. The internet has helped people surround themselves with only voices that agree with them and validate them, often to their detriment. There are entire online communities urging people to cut others out of their lives over the slightest disagre…
Yeah, a lot of it isn't even really about people's choice though. Like, I have so many click-baity things in my YouTube feed even though I ignore almost all of it, because just one click and suddenly that's a thing it's going to show me 10x. I wish there was a place I could go for tech news that just didn't talk about AI. I'm so over it.
I get it, we can use AI agents, and they keep improving. Great, now calm down and go build stuff, and show me the stuff. I really think LLMs and agents are incredible, but I don't want to hear that over and over again for another 5 years.
HN is particularly bad for it right now. Where are all the innovations, cool technical side projects etc happening outside of LLMs?
If you're looking for advice in a situation where you are not sure of what the correct choices are, you will find LLM chat AI to go in circles. It says one thing. Something is dodgy about it, so you raise a tentative objection (as a non-expert). The thing does a "you are completely right, I apologize" about face and then says something different, and things have begun to slide into uncertainty.
That might not exactly be sycophancy, but it's basically the same thing: producing responses that are reflection of what is in the chat, rather than any real shit.
Pick any topic where people disagree. It could be an entirely technical topic in which engineers have settled the questions, and the only contrarians are crackpots. The problem is that the crackpots are out there writing, and this is snarfed into the training data. Crackpots use certain ways of talking about certain subjects. If you use similar vocabulary and concepts that align with some crackpot theory, the AI simply starts predicting tokens according to that, and you are now in crackpot land: what you are saying is validated using the crackpot terms. Next, write in a way that reintroduces rigidity: proper terminology and correct concepts, and, whoa, the AI is an engineer again, contradicting the previous crackpot shit.
I have an alternative view which is that sycophancy erodes trust in AI, from those who are not seeking validation from a machine, but information or advice. If you're looking for advice in a situation where you are not sure of what the correct choices are, you will find LLM chat AI to go in circles. It says one thing. Something is dodgy about it, so you raise a tentative objection (as a non-expert). The thing does a…
I recently built an ai chart product that attempts to add more nuance by injecting extra models to jump in when the first model misses the mark. (http://pellmell.au So far it seems like only a minority of people like it. The main feedback is that answers are confusing or too much to read
> This suggests that people are drawn to AI that unquestioningly validate, even as that validation risks eroding their judgment I believe this is a larger problem than just AI. The internet has helped people surround themselves with only voices that agree with them and validate them, often to their detriment. There are entire online communities urging people to cut others out of their lives over the slightest disagre…
Yeah, a lot of it isn't even really about people's choice though. Like, I have so many click-baity things in my YouTube feed even though I ignore almost all of it, because just one click and suddenly that's a thing it's going to show me 10x. I wish there was a place I could go for tech news that just didn't talk about AI. I'm so over it.
> I wish there was a place I could go for tech news that just didn't talk about AI.
Earlier quoted context omitted.
Except I'm providing evidence.
I'm seeing more examples of you saying you're providing evidence than of you actually providing evidence. So far I've heard a single anecdote about what you did. That's a pretty far cry from the "evidence" you claim to have dropped on multiple comments.
Earlier quoted context omitted.
I think we are noticing it. Are you not? I am not. Congratulations on your side project.
Projects, plural. And weren't you the one who started this complaining of nastiness?