The future belongs to the AI-whisperers – it’s time to be disciplined about your “prompting”!

Animal whisperers are said to:

…practice the art of telepathic animal communication. They also use other intuitive gifts to find out what is really going on with an animal at very deep energetic levels, physically, mentally, emotionally and spiritually.

This language and the concept behind it are a bit over the top, but you get the idea. The notion has been around for a long time. Think Francis of Assisi, and that’s a mere blink of an eye compared with the domestication of animals, which goes back some 20,000 years.

Some Americans of a certain age got their introduction to the term much more recently and directly – from the 1998 film the Horse Whisperer – portraying a guy with a remarkable understanding of horses. (Robert Redford starred; he was then at the peak of his powers and was really whispering throughout the film to a different species and a specific demographic.)

All this came rushing back to mind a week ago when I sat in on a talk on artificial intelligence. The speaker provided an engaging overview; the audience learned a lot. But for me the takeaway was one small piece:

Prompting is a thing:

In artificial intelligence (AI), prompting is the process of communicating with an AI system by providing specific instructions or queries to achieve a desired outcome. Prompts are the interface between human intent and machine output. For example, in Microsoft 365 Copilot, AI prompting involves communicating with the AI model to generate code, content, or responses based on user input.

Wow.

More flashback. Back in those same 1990’s I was talking with a USGS friend about search engines. I mentioned what I was using then, Ask Jeeves (now Ask.com). After a brief, polite silence, my friend gently said: “well my (ten-year-old!) son prefers this search engine called Google.”

Gave it a try – and of course never looked back.

If you started using Google about that time, or earlier, you remember that it (and other search engines) were great for addressing a certain range of questions and problems, but for others, search still required a trip to the library, burrowing into printed word, or consultation with experts. But about every six months, it was necessary to recalibrate, because the material Google had to work with was rapidly growing. Two or three decades later, for many of us, the tables are turned. If we can’t find what we need online, looking into it further isn’t worth the bother; the opportunity cost is simply too great.

Your decades of Google use also taught you something else. You became much more efficient in your queries. You learned what was essential to a good query, and what niceties and frills (such as correct spelling) could be ignored. It’s not just Google and the world’s data bases that have changed. You changed. You learned to whisper to Google.

Back to the present day and artificial intelligence. AI is a higher life form. If Google is a gerbil, AI is that horse.

If you have played around with generative AI you already know that asking the right questions in the right way is everything. And it’s tough. It’s the Google-search problem on steroids. The AI world recognizes this. They know that the productivity boosts that AI can offer improve dramatically as your prompts become more adroit. They also know that if each of us figures out what works totally on our own the process will be too slow.

Online you can find tips. The earlier link provides one such set:

• Be clear: Use plain but clear language.
• Provide context: Provide specifics about who your audience is and what sort of tone you’d like to set.
• Avoid vagueness: Being too vague or broad can lead to generic or irrelevant responses.
• Avoid over-specification: Excessively detailed prompts can confuse the model or limit its creative scope.
• Avoid literal interpretation: AI often interprets prompts literally, so figurative language can lead to unexpected results. 

Hmm. Sounds like rules for talking to another person. And such a single simple list doesn’t cut it. These very general suggestions translate into specifics that are different depending on the use of AI (much as they do depending on whether you’re talking to your life partner, or a work colleague, or a stranger). Accordingly, prompt engineering, or prompting has become a key productivity factor in the AI world:

Prompt engineering is the process of writing natural language text that guides generative AI (artificial intelligence) models to produce desired outputs. The text is called a prompt, and it describes the task the AI should perform. Prompt engineers use creativity and trial and error to create input texts that help the AI interact with users more meaningfully. The goal of prompt engineering is to ensure that AI models produce accurate and relevant outputs.

Coursera, Udemy, and others provide apps and modules on prompt engineering. University computer-science curricula now include coursework in prompt engineering. It’s essentially different depending on the stage of software development. Need AI help in coding? That’s one context for prompt engineering. Want AI help in applications to problems in health care, research, marketing, strategic planning? Effective prompting required there is quite different. And multiple levels of IT between these two ends of the spectrum each require their own prompting skills.

With AI threading through just about every aspect of knowledge work, it’s easy to be dismayed. That’s because AI is not only complicated; it’s rapidly growing in capability. As a result, the nature of all that interlaced professional activity will also be changing. The demands on and desirable attributes of prompting will be changing at the same pace. It won’t be long before AI capabilities are nearly human – or something more. How are humans to keep up? We evolved over hundreds of thousands of years of slower change; our ancestors essentially died in the same world into which they’d be born. The opposite will be true of the present and future generations.

What’s a human to do? Well – fortunately – it turns out that perhaps, instead of whispering, “you should hold your horses.” More in the next post.