𝗜𝘀 𝗣𝗿𝗼𝗺𝗽𝘁 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗗𝗲𝗮𝗱? 𝗠𝘆 𝗥𝗲𝗰𝗲𝗻𝘁 𝗔𝗜 𝗘𝘅𝗽𝗲𝗿𝗶𝗺𝗲𝗻𝘁 𝗦𝗮𝘆𝘀 𝗠𝗮𝘆𝗯𝗲.

Over the past few days, I’ve been working on a side project that left me questioning the very core of prompt engineering.

A friend pitched an idea for a game: players need to find food-related anagrams within a recipe title. He’d need hundreds of recipes and their associated anagrams. Seemed like a perfect use case for AI, right? My goal was to generate a few hundred recipe titles with their corresponding anagrams. I tried using increasingly complex, multi-step prompts, which brought me to the limitations of my prompt engineering skills. The results were disappointing, full of inaccuracies and undesirable results.

Out of frustration, I took a different approach. I asked the LLM (Large Language Model) itself for the prompts necessary to produce the results I wanted. The LLM provided an 8-step prompt sequence which showed me a few steps I had missed. Out of curiosity, I responded, “OK, do those steps,” and the results were astounding. Within seconds, I had 100 accurate recipe titles with their food-related anagrams.

This experience leads me to ponder: Is the skill of prompt engineering dead? If an AI can provide the optimal prompts itself, why would we need to invest the time to become good prompt engineers? I realize that this is provocational. Obviously I had to employ an intermediate level of prompt engineering to receive the correct steps, but it’s clear that the ability to formulate good questions remains an essential skill no matter if we are working with people or AI.

As we stand on the frontier of AI, it’s essential to continuously question and redefine our approaches. This experiment has shown me that sometimes, the best results coming from asking a simple question.

#AI  #Innovation  #ProductLeadership  #PromptEngineering  #FutureOfAI