To engage in prompt engineering, we will need an OpenAI account (since this course focuses on OpenAI) and access to the playground.
Explaining the parameters of OpenAI:
- Model – will determine the parameters available
- Temperature – The higher the temperature, the more variable/creative the responses
- Maximum length – Controls the length of the response (and cost)
- Stop sequences – Stop the response after a certain sequence of characters, such as period or semicolon
- Top P – Controls diversity: 0.5 means it’s considered half the “pool” of available words
- Frequency penalty – Controls likehood of repeating the same line
- Presence penalty – Controls likehood of talking about new topics