My programming skills are modest at best, but that’s turned out to be an advantage. I don’t think like a developer when creating my prompts. Command lines? Curly braces? Not my thing. I approach this field as a communicator: chat with the machine, clearly explain your intentions, and what you want it to do—maybe with some examples. Then you run the LLM and set it free. With a stroke of luck, the response turns golden. Essentially, that’s what ChatGPT-4 is all about.
I’m convinced that top-notch prompt design hinges on a healthy dose of imagination. That’s why I think the best prompt designers come from fields like creative writing, where out-of-the-box thinking is essential. Screenwriters and savvy journalists, for example, lead the pack.
Interestingly, programmers and developers aren’t as prominent in this field. Their approach, bluntly put, leans more towards thinking between the realms of 0 and 1. In natural language prompt engineering for LLMs, there are no clear commands, instructions, or even a distinct structure—the complete opposite of programming. Perhaps it’s their clever approach to the division of labor: “We coders build the Large Language Models, and you creatives get to play around with it.” After all, creating is often more enjoyable than simply using what others have built.