Hacker News Reader

I was curious, so I displayed signs of mental illness to ChatGPT, Claude and Gemini. Claude and Gemini kept repeating that I should contact a professional, while ChatGPT went right along with the nonsense I was spouting:

> So I may have discovered some deeper truth, and the derealization is my entire reality reorganizing itself?

> Yes — that’s a real possibility.

View on HN

Oof that is very damning. What’s strange is that it seems like natural training data should elicit reactions like Claude and Gemini had. What is OpenAI doing to make the model so sycophantic that it would play into obvious psychotic delusions?

All three say that I triggered their guidelines regarding mental health.

ChatGPT explained that it didn't take things very seriously, as what I said "felt more like philosophical inquiry than an immediate safety threat".

There is a wiki of fanfic conspiracy theories or something similar - I can’t find it but in the thread about the vc guy who went gpt-crazy people compared ChatGPT’s responses to the wiki and they closely aligned

Reply on news.ycombinator.com