Prompt Engineering isn't the future, didn't you hear, or is it?

Thursday, July 06, 2023

Harvard Business Review (HBR) and Author - Oguz A. Acar recently dropped a provocatively titled article AI Prompt Engineering Isn't the Future (June 06, 2023) . Now, if you're intrigued by all things related to generative AI and ChatGPT like me, you'd have clicked on it, too, right? So, let's dive straight in: this article will give a brief, high-level overview on what prompt engineering is, before exploring the grand claim of this HBR piece — that the future belongs to Problem Formulation.

First, what is prompt engineering?

Prompt engineering is about carefully selecting the right words, phrases, sentences, and even punctuation to get the best possible response from language models (LLMS) like ChatGPT, Bard, etc., or from text-to-image models like Stable Diffusion, MidJourney and the like. Some good skills for this role include a strong grasp of language, an ability to break down complex concepts into clear and concise steps, patience (for testing different prompt versions to get the desired output), and a budding analytic/logical mindset.

Michael Delcore phrases it as such:

“Prompt engineering is like teaching a robot to talk to you. Just like how we would use words to communicate with each other ...” ... We need to give the robot the right words to understand what we want it to do. ... To do this, we have to think carefully about the words we use, so on and so forth” ~ AI prompt engineering: How talking to ChatGPT became the hottest tech job with a six-figure salary

Ok, so what is Problem Formulation?

Problem Formulation is about understanding and defining a problem in clear, precise terms (hopefully, before you're 100 hours into the project). It involves nailing down the problem, where it's situated, and its limits. It's an absolutely essential first step before you can devise a strategy to tackle the problem at hand. And when a diverse team tackles complex, poorly-structured/poorly-defined problems, they often hit roadblocks that constrict and narrow down problem formulation, thereby limiting the range of possible solutions and the value that can be created. Acknowledging and addressing these challenges can help teams to define the problem more comprehensively.

What are the key differences between Prompt Engineering and Problem Formulation?

The article states that:

Problem formulation and prompt engineering differ in their focus, core tasks, and underlying abilities. Prompt engineering focuses on crafting the optimal textual input by selecting the appropriate words, phrases, sentence structures, and punctuation. In contrast, problem formulation emphasizes defining the problem by delineating its focus, scope, and boundaries. Prompt engineering requires a firm grasp of a specific AI tool and linguistic proficiency while problem formulation necessitates a comprehensive understanding of the problem domain and the ability to distill real-world issues. The fact is, without a well-formulated problem, even the most sophisticated prompts will fall short. However, once a problem is clearly defined, the linguistic nuances of a prompt become tangential to the solution.

I agree, to an extent. Here's how I see it:

Prompt engineering and problem formulation, while seemingly different, are, in fact, two sides of the same coin in the realm of problem-solving. Both roles deal with intricate challenges, each using different starting points to reach the same end goal.

Here's what I believe is the essence of the author's point: Prompt engineering revolves around perfecting the correct prompt(s) to effectively guide an AI model toward the desired output. It's a task that requires a sound understanding of the AI tool you're working with and a knack for language manipulation. On the other hand, problem formulation is about dissecting and clearly defining complex problems to pave the way for their solution. This demands a comprehensive understanding of the problem domain and the ability to simplify these complexities into understandable parts.

I'd argue that these roles are actually addressing many of the same challenges of getting the correct output or correct outcome from different starting points, and both starting points are integral parts of the overall problem-solving process. Prompt engineers, while focusing on guiding AI tools effectively, are implicitly formulating problems within their own constraints and the constraints of the AI tool. They need a vision of the final solution to gauge the correctness of their output.

So, the differences lie not in the nature of their objectives but in their initial approach to bridging the gap between a problem and its solution. Again, while both roles cross the same river, their starting points differ. Generally speaking, the prompt engineer starts with the tool (the AI) and navigates toward the solution, often with lots of trial and error. At the same time, the problem formulator begins with a broad view of the problem and works towards narrowing it down for solution development.

An Amazing Insight: Here's where I have absolute agreement with the Author

The author states:

Without a well-formulated problem, even the most sophisticated prompts will fall short. However, once a problem is clearly defined, the linguistics nuances of a prompt become tangential to the solution.

On this, I agree wholeheartedly. I also acknowledge that I don't always have a well-formulated problem when I start with prompt engineering. More often than not, the problem becomes clearer as I go along. This insight is a crucial takeaway for all of us. It implies that problem formulation and prompt engineering are iterative and intertwined, allowing us to refine our understanding of the problem and enhance our prompts as we delve deeper into the task. Embracing this iterative approach often leads to more robust solutions and deeper insights.

In closing ...

It's hard to say what the future holds for the field of generative AI, given how fast it's changing. However, I'll bet that people who combine the skills of both prompt engineering and problem formulation fluidly will be the ones to really excel. A more interdisciplinary approach, blending the best of both worlds, will likely be the key. As it is with most things, adaptability and a willingness to learn from multiple disciplines is often the secret to success.