Prompt engineering is all the rage these days, with companies and individuals claiming expertise in crafting the perfect prompts to get chatbots like Claude to generate high-quality content. But here’s a dirty little secret – most people saying they are prompt experts are posers and liars.
As someone with a background in architecture and web design in the early days of the internet, I’ve seen this pattern before. When Computer Aided Design (CAD) first emerged commercially, companies had to decide whether to train existing architects on the new software or hire CAD experts and teach them architecture.

Should we hire a CAD guy and teach him how to be an Architect –or– Should we hire an Architect and teach him how to be a CAD guy?

Although AI was in the labs at the time it is very pertinent to note that CAD transformed the world far earlier than AI because CAD automated the way people make drawings required to design and build just about everything.

The world-changing transformation was internal however; in architectural and engineering offices and within manufacturing plants world-wide. CAD did not take on a public persona to become known and understood by the general public and to this day some people do not know what the CAD acronym represents.

Generative AI on the other hand has during 2023 created a world-wide phenomena; few do not know that AI is the acronym representing Artificial Intelligence.

Like a Tsunami, a wave of automation made possible by AI is sweeping away the old and we find the same question asked during earlier transformative periods of change being asked once again:

Should we hire an AI guy and teach him how to be a (whatever) –or– Should we hire a (whatever) and teach him how to be an AI guy?

The dilemma prevails across time and each time we find the context of the answer constrained to that subject matter expertise we develop in the world where we perform human activities such as “going to work.

The context of the dilemma as it pertains to generative AI is one which entails using human faculties to engage in conversations with machines.

Humans literally speaking to and with machines are in the labs in every nation of the world. For now we are taking baby steps by submitting text to a machine that is a type of computer programming using the same language we use to speak and write on a daily basis; no matter which nation or part of the world we are from.

For the time being “conversation” with machines typically refers to using our written language to ask a machine to tell us what we want to know. That process is called “prompt engineering.”

Hence the pervasive existence of the “should we” paradox.

As most who have tried the conversational bots have discovered:

Generative AI is like a dog; give it a biscuit and it will roll over, do a trick and then wait for another biscuit.

What that means is even the simplest prompts submitted to a machine will produce a response of some kind. The reality is though:

Effective prompt engineering requires deep knowledge in as many areas of application as possible –and– a mastery of the language one uses to speak and write in every day life; everything about the use of the language structurally and academically.

Prompt Engineering is –computer programming– with the same conversational natural language humans speak and write with every day.

Garbage In – Garbage Out.

If you want your dog to go out to find the newspaper, find your slippers and your pipe and bring them to you and then take your shoes off, put your slippers onto your feet, put the newspaper on your lap, fill your pipe with your favorite smoking material and light it for you then go put your shoes away in your bed room so they are ready to be worn tomorrow morning you damn well better hire an Architect and teach him how to be CAD guy so he knows –exactly– what needs to be known and said to the machine if you expect the dog to do the trick you –REALLY– want.

You need expertise in the niche you’re targeting, whether that’s architecture, medicine, or poetry. You need rhetorical skills to speak the language models understand. And you need to understand the technical side – linguistics, computational linguistics, philosophy, etc. Prompting isn’t just about stringing words together, it’s about using logic and reasoning to instruct computing machines how to function in our behalf.

Prompt Engineering Is –Computer Programming– With The Same Conversational Natural Language Humans Speak And Write With Every Day.

So when looking for help with prompts, don’t just find someone claiming prompt expertise. Ensure that someone understands your field of expertise and has extraordinary communication skills using your native language, mine being English for example.

The best results when prompt engineering combines subject matter mastery with linguistics, rhetoric, philosophical reasoning ability e.g. language skills that can be used in a programmatic and an orderly way to describe an optimal output.

Expertise in prompting alone is marginally useful without domain knowledge. Expertise in domain knowledge alone has become just as  marginally useful.