The concept of a prompt engineering career has received attention with the explosion of generative AI tools and platforms. A prompt engineer is effectively a “prompt whisperer” who knows the ins and outs of creating prompts that will be effective in getting AI to return results that match the need. I will outline here why I think that it would be wise for readers to avoid betting their career on prompt engineering. If you disagree, feel free to comment and debate!
What Prompt Engineers Address
One challenge with all types of generative AI is figuring out how to ask for what you want in a way that causes the response to match your intent. For example, suppose you want an image of five zombies in a rowboat in the middle of a conference room debating a data science trend. What is the best way to construct your prompt to get images back that are close to what you’re looking for? Are you looking for a realistic or cartoon feel? Any specific color schemes? What style of zombies? While there are myriad nuances that can be added to a prompt, which are most important for the goal at hand? How do you structure the prompt to maximize the match between the output and your goals?
One answer to this challenge is the proposed role of “prompt engineer” — a specialist that has expertise in manipulating prompts in often subtle and sophisticated ways to generate acceptable responses much faster than the average person. Naturally, we’d all be happy to have a prompt engineer at the ready to accelerate our success with generative AI. However, there are some practical realities that will dampen the ability to turn that enthusiasm into a viable long-term role.
Practical Issue 1: Cost / Benefit
A major practical constraint on the viability of prompt engineering as a career is cost / benefit analysis. No prompt engineers existed even a year ago and very few, if any, companies have one in place today. Thus, each prompt engineer will be a net new budget expense that will have to be justified based on the value that a prompt engineer can provide.
The value of a prompt engineer is tied to an ability to create an increased quality of prompt output while simultaneously generating them at a faster pace. However, it will take a lot of prompt requests to overcome the cost of the employee. If it takes me 10 minutes to get an acceptable response from generative AI and a prompt engineer can do it in 5 minutes, that’s a 50% savings. That sounds great!
But to justify the cost, think about how many times that must be repeated to cover the cost and/or how much better quality the responses must be. It isn’t impossible to make the numbers work, but the numbers probably won’t work for more than a few prompt engineers at even a very large company. Hence, few prompt engineer roles and limited career options.
Practical Issue 2: Organic Experience And Learned Expertise
Most of us aren’t that great at creating prompts for generative AI applications today. Then again, most of us weren’t initially great at creating classic search queries either. However, as people began to interact with search engines routinely, most of us got good enough to get by. We never hear of the need for “search engineers” to help us with our searches because most of us are good enough that it isn’t worth extra cost to have someone help us. The same will happen with generative AI prompting.
In addition to each of us getting better at creating our prompts, the generative AI engines will get better at interpreting our prompts. After all, the engine developers have an incentive to make them as responsive and accurate as possible. Combining the increased skills of users with the increased power of AI engines will make prompt engineers even harder to justify.
Practical Issue 3: Who Wants The Job?
Historically, roles like help desk support are hard to staff. The people who really know their stuff and are expert enough to answer a broad range of user questions will often want to apply their skills in a way outside of answering questions at a help desk. So, while there are some really good help desk people, it is hard in practice to fully deliver on the goal of ubiquitous and deep expertise.
Similarly, someone who is good enough prompting generative AI to become a prompt engineer often won’t want to do it for a living. Let’s be honest … how many people will find it exciting to sit day after day and wait for the next order from a user to create a prompt? Hence, many of the most qualified people won’t pursue the role, making it hard to staff with consistently high quality … and the less effective the person in the role, the less likely they’ll overcome practical issues #1 and #2.
Summary
Putting it all together, my conclusion is that prompt engineering is going nowhere as a career path. Most people won’t need much support as they gain more experience and expertise, it will be hard for a prompt engineer to add enough incremental prompt efficiency and quality to warrant the costs, and those who are best at it will often prefer another job.
Perhaps there will be a few prompt engineers in place at some of the largest organizations that make significant use of generative AI. However, the numbers will be limited, and I wouldn’t recommend betting your career on the field. I simply don’t see prompt engineers becoming a thing.
Originally posted in the Analytics Matters newsletter on LinkedIn
The post Why Prompt Engineering Won’t Be A Thing appeared first on Datafloq.
