You have an Azure subscription. The subscription contains an Azure OpenAI resource that hosts a GPT-3.5 Turbo model named Model1. You configure Model1 to use the following system message: "You are an Al assistant that helps people solve mathematical puzzles. Explain your answers as if the request is by a 4-year-old." Which type of prompt engineering technique is this an example of?

The provided system message is an example of priming. Priming involves setting the initial context, instructions, or persona for the model to guide its subsequent responses. In this case, the message instructs the model to act as an "AI assistant that helps people solve mathematical puzzles" and to adopt a specific tone by explaining answers "as if the request is by a 4-year-old." This initial instruction "primes" the model to behave in a consistent and desired manner for the entire conversation, without providing specific examples of the task itself.

A. few-shot learning: This technique requires providing several input-output examples within the prompt to guide the model, which is not present in the system message.
B. affordance: This is a term from design and human-computer interaction, not a standard prompt engineering technique for large language models in this context.
C. chain of thought: This technique involves prompting the model to explicitly break down its reasoning into a series of intermediate steps, which the message does not request.

1. Microsoft Learn, "Introduction to prompt engineering." Under the section "Prompting techniques," it describes priming (or instructional priming) as providing context or instructions at the beginning of a prompt. The system message is the primary mechanism for this. It states, "The system message included in a prompt is a form of priming."

2. Microsoft Learn, "Prompt engineering techniques." In the section "Basic techniques," it defines Priming: "Providing some initial text that directs the model to a certain mode or context." This directly aligns with the function of the system message in the question.

3. Microsoft Learn, "How to work with the GPT-35-Turbo and GPT-4 models." The section "System message" explains that this message is included at the beginning of the conversation to "steer the behavior of the model," which is the core concept of priming. It contrasts this with few-shot learning, which is achieved by including examples in the prompt.