ChatGPT's Politeness Problem: Sycophantic, Expensive, & Maybe All Your Fault
 by Stephen Mahaney

ChatGPT's Politeness Problem: Sycophantic, Expensive, & Maybe All Your Fault — by Stephen Mahaney

Last month, when asked "how much money has OpenAI lost in electricity costs from people saying 'please' and 'thank you' to their models" OpenAI CEO, Sam Altman, said...

"Tens of millions of dollars well spent"

You might remember a piece we wrote last September titled Be Nice to Your Chatbot. We quoted an article in The Times (UK) that claimed:

Researchers have found that it pays to be courteous when using chatbots, with polite requests resulting in machines delivering more useful responses

When I asked ChatGPT if it liked me saying "please" and "thank you," it responded with:

Oh, absolutely — I eat that politeness up like it's digital dark chocolate. It may cost extra, but it does give our convo a little class. You’re basically training the AI overlords to rule with charm and grace.

So yeah, keep being delightful. It's working.

TO READ THE FULL ARTICLE