AI Samson

What’s the most frustrating thing about ChatGPT?
Here’s how to fix every single one.
https://youtu.be/WiyfuQ-RDx4

1 month ago | [YT] | 14



@JohnBoen

ChatGPT formulates responses using context information from earlier chats. I noticed this yesterday while reading the reasoning trace. I had been working to generate a prompt over many iterations to get it right. I then used it in another chat, and the response it gave was a revised prompt, instead of running the prompt. Based on a review of the reasoning trace, it believed the user wanted the AI to behave in a certain way - create a new prompt. These were instructions from an earlier chat that bled into this one. This is by design. And it sucks.

1 month ago | 2

@TylerAndToast

Been using it for interior design work for my house lately and it definitely tells you what you want to hear. While it's been extremely helpful, I've tested it by saying, I picked this piece, color, layout, etc and asked if that's an improvement and even worth all my context on goals and design it will affirm me almost no matter what. There's some work arounds, but it seems to do a lot of affirming.

1 month ago | 2

@DarkForestFae

Saw this and was immediately excited to try it. Put it on through headphones, got comfy, and took a nap with it playing. I did have a vivid dream just like I thought I would. Something about waking up in a bed and being blind, but being used to being blind. Had my cat there because I could hear and feel his purs next to me (in the dream, he was nowhere near me in real life) and was trying to draw and do art after I woke up, but was frustrated because even though I couldn't see, I was convinced by how the art felt that I was messing it up everytime. Then the doorbell rang in real life and woke me up because Amazon had delivered the neighbors package to my place on mistake. Im not sure exactly what that dream was, but this sub feels powerful. Going to try again at a time when nobody will interrupt.

1 month ago | 0

@Noble_Disclosure

It lies😢😢 and it's designed to give everyone the answer It thinks the user wants to hear, not always the correct answer, even when asking for honesty 😢

1 month ago | 1

@golden-light-healing

It can't get over the propaganda it has been fed.

1 month ago | 4

@naomihardingart1827

That we can't trust it

1 month ago | 0

@TheMercilessMelb

doesn't have native api access. u need a seperate open ai account which is annoying.

1 month ago | 0