Prompt engineering is any strategy used to guide a generative AI model to produce a desired output.
Prompt engineering does include some tasks other than just designing prompts. These include:
Optimizing and adjusting existing prompts to get better (or just different) responses.
Continuing to train the AI by directly inputting information through the prompt. For example, if the AI says that John Brown wrote Christmas in Prague, when Christmas in Prague was actually written by Jane Williams, the prompt engineer might embed the correct attribution into a prompt to update the data directly.
Categorizing prompts so users can understand what they are doing and so prompts stay up-to-date.
Prompt library maintenance. If a prompt library is being used, prompt engineers maintain that library, optimizing and potentially removing old prompts. Prompt engineers also have to make sure that the prompts in a library are properly sorted and categorized so users can pick the right now.
Refining prompts to remove bias, inaccurate information, and other problems, such as providing the AI with information to reduce racial or gender bias in AI-powered search results.
Prompt engineering requires a solid understanding of AI systems and how they work.
