I’ve said it before, and I’ll say it again. Prompt Engineering is teetering on the edge of irrelevancy. Writing prompts in ChatGPT or Midjourney or [insert your favorite AI tool] in order to interact and get work from them has a high chance of becoming obsolete as a human-AI interface.
AutoGPT and Agent AIs run much simpler. Users input a mission statement or objective and some goals, and the Agent AI will communicate with LLMs to get the work done. So all of the formatting and descriptive language separating a good prompt from a bad one won’t be a necessary “skill” to learn.
That’s the long-term risk to Prompt Engineering. AutoGPT and Agent AIs are not better than ChatGPT today in terms of output and impact. However, they hold the promise of an easier user experience.
The short-term risk to Prompt Engineering will come from prompt-generating LLMs.
The reality is that every day, we create billions of AI prompts. I couldn’t find an exact figure, but I imagine it’s in the billions. These prompts could then be used to fine-tune an AI on how to generate effective prompts.
It’s a meta concept to think about. Training an AI on how to communicate with AI in order to get better outputs from the AI. But that’s exactly what tools like PromptStorm are, in essence.
Prompt Storm is a Google Chrome extension that generates well-crafted prompts for ChatGPT in just a few clicks and by adding just a bit of information.
As you’ll see in the series of screenshots walking through using the tool, it reduces at least 90% of the prompt-writing process. And with a little vision, you will see how tools like this will do away with (most) prompt writing.
The Prompt Storm plugin, once installed, appears in the upper-right corner of ChatGPT:
Once opened, you first pick from a handful of general categories (or you can search for prompts):
Then you choose a subcategory:
Next, you select a prompt from their database in that category and subcategory:
That will generate and populate the prompt in the chatbox:
Then you describe the topic and number of outputs; both will auto-populate in the prompt in the chatbox:
Press enter and you get your results:
Tools like this will only get more sophisticated, expansive, and accurate. This is how prompt engineering will change in the short-term.