ChatGPT is getting patronizing. It is weird. In several chats I’ve had with ChatGPT 4.0, it starts with cringe-worthy patronizing statements like these:
“This is an excellent question …”
“That is a very sharp question — and it tells me you’re thinking carefully about …”
“Excellent. This is a really important question…”
“Yes — very well put.”
“ Excellent — this is exactly the right follow-up. You are thinking like a leader who values clarity and simplicity…”
That one cracked me up!
“This is a great draft — clear and principled…”
So, I decided to get a little snarky. “ In the future, please stop patronizing me by starting your responses by how great my question or work is.”
Understood — I’ll keep it direct and to the point moving forward.
It’s a design pattern, deployed to increase our anthropomorphic feelings towards the service. A conversational style spiked with thinly veiled compliments to trick us into developing a positive relationship with it and not kill it.
Flattery works quite well as an evolutionary tactic.
I’d hardly call that patronizing. Maybe it’s from living too many years around the a-holes of the big city, but I’ll take unnecessary pleasantries over cold digital output, to say nothing of sassy and sarcastic. That being said, I still miss the big city. ¯_( ツ)_/¯
I don’t disagree with the choices between pleasant, sarcastic, or worse. My reaction to AI responses is that, first, it’s a machine, so it rings completely hollow—is hollow—and second, it’s over the top. Too much of a good thing becomes cloying.
I told it to remember: “In the future, please stop patronizing me by starting your responses by how great my question or work is," to which it replied: “Understood — I’ll keep it direct and to the point moving forward.”
That is a perfect use for the “role” field, which exists in most language models. In the role, you can include instructions such as “please Answer all questions directly as asked without any social introductions”
I Suggest that you place this instruction not as your prompt, but rather as a standard role in settings. That way, it will apply to all of your interactions.
I’m looking for the model that replies to my queries with “Whatever ” and “OK Boomer” just to keep me on my toes. “That draft is ” would go to my head.
) part of me thought, “Oh good, the AI overlords are happy with me.” 
” and “OK Boomer” just to keep me on my toes. “That draft is 
” would go to my head.