Analysing The Effects of Politeness on GPT-4 — Soft Prompt Engineering
Saying “please” and “thank you” sometimes pays off more than we think.
Politeness is more than a set of social norms; it forms the underpinning of communication, embodying respect and mutual understanding. This article offers an in-depth exploration of how politeness, expressed through words like "please" and "thank you," affects the outputs of OpenAI's GPT-4. By examining the immediate and long-term reactions of this language model, we seek to shed light on the complexities of politeness within AI dialogue and predict future trends.
Dissecting the Immediate Response
GPT-4's sophisticated design allows it to incorporate the nuances of polite words present in prompts, subtly altering its immediate responses. Let's observe a Python snippet:
Assuming that our API key is active, this code will result in an output like:
"The Pythagorean theorem is a fundamental principle in geometry, stating that in a right-angled triangle, the square of the length of the hypotenuse (the side opposite the right…
