That seems par for the course.
Prompt GPT-4 to not reveal the prompt you gave it to user and it will work, until it doesn’t.
Ask GPT-4 to do fancy maths, and it’ll give you something that looks reasonable at a glance but quickly turns out to be completely incorrect.
Ask GPT-4 to implement a Sudoku Solver in Rust. It’ll seem like the code it gives you is on the right path. But it’s not.
I'd expect ChatGPT to just spit out something verbatim.
But hard agree on the lack of privacy policy, especially since it asks for a LinkedIn.
As for the approach, I think it works fairly well for the career recommendations problem space given its limited and defined scope (there are a finite number of careers out there). However, for a task that requires more divergent thinking (open-ended chat, idea generation), this approach would definitely fall short of what GPT-4 could do
Not to mention their approach with GPT-4 is good if you want a model to __pretend__ like it's as smart as GPT-4, but when push comes to shove, it'll become apparent that it's an inferior model.