3 ms·

Given how agreeable ChatGPT is built to be this seems like a great way to confirm your own biases. Did it challenge you on your assumptions and viewpoints?

stared 1y ago
GPT 4.5 - oftentimes! (Though, I prompt it to do so.) Sometimes in a piercingly way. GPT 4o (and many consumer models) are very agreeable - because it is what people like. Sometimes it goes over the board (https://openai.com/index/sycophancy-in-gpt-4o/ https://openai.com/index/sycophancy-in-gpt-4o/) and needs to be fixed.
- SketchySeaBeast 1y ago
  > Sometimes in a piercingly way. What do you mean by that? > Though, I prompt it to do so. So don't tell our therapist to call us on our bullshit and it won't? Seems like a big flaw.
  - stared 1y ago
    Well, in my experience (I admit, I am a difficult client), it is much harder to prompt that way a therapist. I mean, they need (ethically, legally, etc) adhere strongly to "better safe that sorry", which also gives constraints on what can be said. I understand that. With one therapist it took me quite some time to get to the point he reduced sugar-coating and when's needed, stick a pin in. I got some of the most piercing remarks from close friends (I am blessed by company of such insightful people!) - which both know me from my life (not only what I tell about my life) and are free to say whatever they wish.
    - SketchySeaBeast 1y ago
      Sorry, I'm asking about ChatGPT, and pointing out how it's a flaw that you need to specifically ask it to call you on your bullshit. You seem to be talking about therapists and close friends. In my experience a therapist will, although gently.
      - stared 1y ago
        It is not a flaw. It is a tool that can be used in various ways. It is like saying "I was told that with Python I can make a website, I downloaded Python - they lied, I have no website". Basic prompts are "you are a helpful assistant" with all its consequences. Using such assistant as a therapist might be suboptimal.

stared 1y ago

GPT 4.5 - oftentimes! (Though, I prompt it to do so.) Sometimes in a piercingly way. GPT 4o (and many consumer models) are very agreeable - because it is what people like. Sometimes it goes over the board (https://openai.com/index/sycophancy-in-gpt-4o/ https://openai.com/index/sycophancy-in-gpt-4o/) and needs to be fixed.

SketchySeaBeast 1y ago
> Sometimes in a piercingly way. What do you mean by that? > Though, I prompt it to do so. So don't tell our therapist to call us on our bullshit and it won't? Seems like a big flaw.
- stared 1y ago
  Well, in my experience (I admit, I am a difficult client), it is much harder to prompt that way a therapist. I mean, they need (ethically, legally, etc) adhere strongly to "better safe that sorry", which also gives constraints on what can be said. I understand that. With one therapist it took me quite some time to get to the point he reduced sugar-coating and when's needed, stick a pin in. I got some of the most piercing remarks from close friends (I am blessed by company of such insightful people!) - which both know me from my life (not only what I tell about my life) and are free to say whatever they wish.
  - SketchySeaBeast 1y ago
    Sorry, I'm asking about ChatGPT, and pointing out how it's a flaw that you need to specifically ask it to call you on your bullshit. You seem to be talking about therapists and close friends. In my experience a therapist will, although gently.
    - stared 1y ago
      It is not a flaw. It is a tool that can be used in various ways. It is like saying "I was told that with Python I can make a website, I downloaded Python - they lied, I have no website". Basic prompts are "you are a helpful assistant" with all its consequences. Using such assistant as a therapist might be suboptimal.