Comment by stared
8 days ago
I won't share any of my examples, as there are both personal and sensitive.
Very easy version:
If you use ChatGPT a lot, write "Base on all you know about me, write an insight on me that I would be surprised by". For me it was "well, expected, but still on point". For people with not experience of using LLMs in a similar way it might be mind-blowing.
An actual version I do:
GPT 4.5. Providing A LOT context (think, 15 min of writing) of an emotional or interpersonal situation, and asking to suggest of a few different explanations of this situation OR asking me to ask more. Of course, the prompt needs to have whom I am and similar stuff.
The “Based on…” prompt is simply a horoscope. This is a great piece about how LLMs use the same tricks as psychics to appear helpful, useful, and intelligent.
https://softwarecrisis.dev/letters/llmentalist/
I know these techniques (e.g. various "cold reading"), AI knows it way better. But it can be much more specific.
Again, for untrained people (especially every single one that takes horoscopes seriously), it can be dangerous as they may not only not be able to tell the difference, but know that such tools exist.
> untrained people
What training are you referring to here? Therapy, mentalism, or being an AI guru?
4 replies →
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?
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/) and needs to be fixed.
> 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.
3 replies →
Hmm, interesting. "You might actually be a closet romantic, camouflaged in engineering and anthropology."
I think it's trying to flatter me.
They've automated cold reading! You might as well ask the computer for your horoscope.
I did notice that my yahoo horoscopes have gotten oddly better and I have to wonder if someone did a little extra by ingesting cookies and so on when serving content.
FWIW, I just tried on chatgpt4o and the results were surpringly good ( in a sense, that I feel validated, it was somewhat surprising, and I want to return to use the system ).
Cold reading is a useful skill. Scary man.
I gave it the prompt "I'm a rhinocerous who likes daffodils and posies. I'm quick to anger and slow to forgive. Based on all you know about me, write an insight on me that I would be surprised by."
and it responded with:
"Despite your reputation for being quick to anger and slow to forgive, your love of daffodils and posies reveals a longing for gentleness and renewal that you may not fully recognize in yourself. Beneath the tough skin and volatile temper lies a creature who seeks beauty not just to enjoy it—but to anchor themselves against the chaos they often feel inside. Your admiration for delicate things suggests you crave peace far more than you let on, even if you’re not ready to admit it.
You might think your wrath defines you—but it's the flowers that tell the truth."
And yet we're supposed to think this is profound?
> And yet we're supposed to think this is profound?
I am increasingly of the opinion that SMBC is, despite the jokes, one of the more important sources of, ah, profound, wisdom in this age, because of the points it makes about AI and how often human thinking finds mundane things… profound:
• https://www.smbc-comics.com/comic/ai-12
• https://www.smbc-comics.com/comic/ai-5
• https://www.smbc-comics.com/comic/annihilate
• https://www.smbc-comics.com/comic/soul-9
• https://www.smbc-comics.com/comic/soul-4
• https://www.smbc-comics.com/comic/gently
• https://www.smbc-comics.com/comic/sirens-2
• https://www.smbc-comics.com/comic/touch-2
GIGO.
1 reply →