Is prompt engineering a skill or a job?
The future belongs to those who can generate good prompts, but does that mean it's a job?
Back in the late ‘90s and early 2000s, there was a cottage industry of search engine specialists who devised arcane Boolean searches to be used with Google. But as Google’s search technology improved, and Google focused its attention on serving a large audience to advertisers, Boolean searchers became, if not quite deprecated, less of a focus for Google. Arcane Boolean searches were great for people searching for information in the tails. But Google realized that there was no money in serving the interests of the tails and re-oriented its business, and so its search results, around the mass middle. A consequence of this is that normies, i.e., the mass market, could more easily find what they were searching for, without having to learn arcane Boolean syntax. And, with that pivot, a cottage industry died.
Thanks for reading Buy the Rumor; Sell the News! Subscribe for free to receive new posts and support my work.
You can, of course, still use Boolean strings for Google searches, but no one is going to pay you for your search engineering. To the extent that skillfully using Google is a requirement for a job today, it is a skill that is considered table stakes. No one hires for a search engineer to devise Boolean searches for a company.
I suspect that prompt engineering will follow a similar course. This is to say, I expect that now, and for the next few years, there will be some career alpha to developing skills as a prompt engineer, but, over time, the skill will perfuse through white collar knowledge work, and become table stakes. Further, technology tends to abstract complexity away over time. How large language models work may be inscrutable, even to researchers, but one need not know how LLMs work any more than did someone in 2001 need to know how backrub worked in order to invoke complex Boolean searches.
I asked BingAI whether prompt engineering is a skill or a job, and this is its overly diplomatic, though fluent, output:
Prompt engineering is a concept in artificial intelligence, particularly natural language processing (NLP), that involves designing effective inputs for AI models to perform various tasks. For example, if we want an AI model to generate a summary of a text, we can use a prompt like “Summarize the following text in one sentence:” followed by the text. The prompt acts as an instruction for the AI model to understand and execute the task.
Prompt engineering is both a skill and a job. It is a skill because it requires creativity, logic and domain knowledge to craft prompts that can elicit high-quality outputs from AI models. It also requires experimentation and evaluation to fine-tine and optimize the prompts for different tasks and domains. Prompt engineering is not just about writing questions or phrases, but about understanding how AI models work and how they can be guided to produce desired results.
Prompt engineering is also a job because it has become an essential part of many applications that use AI models, such as marketing campaigns, emails, ads, chatbots and more. Prompt engineers are professionals who specialize in creating prompts for various purposes and audiences. They work with AI developers, data scientists and domain experts to design promtps that can leverage the capabilities of AI models while avoiding their limitations and biases. Prompt engineers are also responsible for testing and validating the outputs generated by the prompts and ensuring they meet the quality standards and expectations of the users.
Prompt engineering is therefore both a skill and a job that has emerged as a result of the rapid development of artificial intellignece. It is an important aspect of using AI models effectively and responsibly in various domains and applications.
The footnotes in the block quote were inserted by Bing. I don’t know why footnotes 2 and 3 are together, when one could have sufficed.
Though Bing’s output is factually correct (and impressive!) I do not think it follows that prompt engineering will always be considered a discrete job. Bing’s reponse seems to be infected with the same recency bias that I see people fall victim to. That is to say, people look at something like Anthropic’s “prompt engineer” job posting and assume that it is the future, as opposed to it being a mere artifact of our current time.