What Is a Prompt and Why Is It Important?

In the context of generative AI, a prompt refers to the input the user provides to the generative models. When we talk about ChatGPT and OpenAI models, prompts mainly refer to a piece of text in natural language or—as we saw for Codex models—programming languages.

Importance of well-defined prompts

Prompts are the only way users can control the output generated by those models. As such, there is no surprise in saying that the quality of the prompts used to engage the AI system in a conversation is key to determining its success.

Well-defined prompts are the recipe for a successful conversation that covers the topics of interest; poorly-defined prompts, on the other hand, not only prevent the conversation from being useful to the user but could also lead to potentially misleading content. As we saw in the previous section, ChatGPT still suffers from some limitations, such as the 2022 cut-off of its knowledge base, as well as another caveat typical of generative AI systems in general, such as the phenomenon of hallucination.

Example of a poorly defined prompt

Imagine that we want to learn about convolutional neural networks (CNNs). Since we have no knowledge about artificial neural networks (ANNs), we need to start with a gentle introduction. The idea is that starting from the first output, we can deep-dive into relevant elements.

Let’s examine the result of a poorly-defined prompt in ChatGPT: