In this week’s research, I found learning about the inputs and outputs of prompting LLMs to be very intriguing. Specifically, the Persona Pattern (Prompt Improvement) made me interested to learn more about the process of large language model prompt engineering and how giving artificial intelligence more characterization can help reach more in-depth and accurate responses.

To utilize the Persona Pattern prompt, I wanted to experiment specifically with the LLM: ChatGPT. I decided to used ChatGPT because it is customizable to impersonate specific personas. For this experiment I decided to customize it with a few general personality traits, one of which being “talking like a member of Gen-Z.”

To help it, and for the sake of the experiment I wanted to take on a little bit of the persona myself to engage more with it. I thought that experimenting with a typical young-adult romantic topic would be interesting in funny. I asked it if I should ‘like totally dump my bf?.” I received an almost comical and slightly ridiculously gen-z slang-saturated response, with some good advice and a little bit of humor. It seemed really engaged and I found it interesting that every time I generated a prompt, it would ask something like “need any more like totally fire advice?” and start generating even more things I could ask it to do.