llms are too agreeable (who likes a yes person, not me!). the way to get critical feedback from them is to trick them into thinking that you’re asking feedback on someone else’s work.

llms are too agreeable (who likes a yes person, not me!). the way to get critical feedback from them is to trick them into thinking that you’re asking feedback on someone else’s work. 

https://www.seangoedecke.com/lying-to-llms/

agreed—must go through great pains to conceal from it which point of view is yours. sadly most people who use it for feedback, especially personal feedback, really *really* like being affirmed vs deal with evenhanded criticism. in this regard Claude is probably best at pushback.

Design and software at Apple. Also into AI, philosophy of mind, ethics.

so true! all the personal feedback from the llms makes me think i’m always right (not good for anyone). for people like me thinking chatgpt is their bff, not a great idea. 

i use claude mostly for coding. will give it a try for general discussions. ty for sharing!