ChatGPT’s tendency to be overly supportive and encouraging is eating your brain alive.
A ChatGPT session is an echo chamber to end all other echo chambers — it’s just you, an overly friendly AI, and all your thoughts, dreams, desires, and secrets endlessly affirmed, validated, and supported.
Why is this dangerous? Well, like any feedback loop, it becomes vicious. One day you’re casually brainstorming some ideas with ChatGPT, and the next you’re sucked into a delusion of grandeur.
Maybe I’m being a little dramatic here. However, even on a small scale, this feedback loop can be problematic, especially if you’re like me and use LLMs daily.
I’m writing this article to help you be as conscious of this problem as possible, and to give you some prompts to avoid it.
This “yes man” problem has existed since the birth of ChatGPT, but it was brought to the public eye at large last week when a simple platform-wide system prompt update from OpenAI made it so much worse.