Prompt Engineer is the New Job of the Future? — I don’t think so
When talking about “jobs of the future” in the age of AI, the conversation often comes across AI Specialists, Data Scientists, and Prompt Engineers. So far so good. The appearance of the first two professions in this list are hard to dispute, since the demand for (good) AI solutions also requires people with the necessary knowledge to develop them. But what about the Prompt Engineer?
What does a Prompt Engineer do?
At first glance, the thought that Prompt Engineers should be part of the list is understandable. Artificial intelligence can save many hours of strenuous mental work if you ask it the right questions in the right way. If you ask the wrong questions, you won’t get the desired result. If you ask the right questions in the wrong way, the result is probably of poor quality. Let’s take this article as an example. Normally, I write a draft text and ask the AI to create a title based on it, since I’m terrible at writing titles :(. Fot this, I could proceed as follows..
Providing the wrong task
Prompt: “Summarize the following article in one short sentence: In the age of […]”.
Output from GPT: The article explores the role of a Prompt Engineer in optimizing questions for AI and suggests that anyone can potentially become one with the right tools and skills.
Why is this wrong?: A title is not a summary! It should generate interest in a topic and should not give away the punchline. A summary does not meet these goals.
Providing the right task asked in the wrong way
Prompt: “Recommend a title for the following article: In the age of […]”.
Output from GPT: “Unleashing the AI Revolution: The Secrets of Prompt Engineering Revealed”
Why is this wrong?: This title is really not exciting and would get lost among the multitude of interesting articles on Medium. Also: wrong punchline :D
Providing the right task asked in the right way
Prompt: “You are an author for an article on the Medium platform. First summarize the punchlines of this article. Then create a list of concise and provocative titles for the following article. Keep in mind that your target audience already has a good knowledge of artificial intelligence. Here is the article: In the age of […]”
Output from GPT: Prompt Engineering: A Dying Art or a Transforming Profession?
Why is this right?: Here, GPT-4 is first asked to reflect the punchline of the article and then fed with the necessary context and the quality criteria important to the author are provided.
So, a Prompt Engineer is a specialist who knows which questions to ask — let’s call it the “What-Skill” — and understands his AI tools well enough to know how to formulate a question — the “How-Skill”. Through his well-thought-out questions, he can unlock the full potential of AI; an ability that is still unreachable for the average user today. The crazy thing: according to a Forbes Article, this job can lead to incredible salaries up to 335k$/y. Sure, because after all, with the help of AI, tasks can be solved that previously required entire teams; suddenly, all you need is one person experienced in prompting. But why isn’t it a so called “Job of the Future”? Let’s deep dive..
Either everyone is a Prompt Engineer, or no one
.. and start with the What-Skill.
The ability to formulate the right tasks is related to a person’s expertise. Only a journalist understands what makes a good title in order to delegate it as a task to an AI. A designer has the knowledge to specify a style before generating an image; a developer specifies interfaces and processes before delegating automation. A Prompt Engineer though, without the necessary expertise, is constantly dependent on the assistance of experts to provide the What-Skill and only stands out because of his How-Skill. He simply knows better how to input a task into an AI.. for now.
On the internet today, there are already numerous instructions for good prompts and a whole lot of lists with thousands of useful prompting examples.
Technically, it is not that difficult to train an AI to use these prompting rules and thus ensure better outputs. And since these “How-Skills” are linguistic skills, they are ideally suited to be imitated by language models like ChatGPT. For example, you can already improve your prompt today with the help of the Prompt Perfect Plugin for GPT-4 (so far only usable with the paid version) and thus automate the task of a Prompt Engineer without deeper knowledge of AI. Let’s try it with the title example from the previous chapter.
Prompt: “Recommend a title for the following article perfect: In the age of […] ”.
Output from Prompt Perfect: What would be the most appropriate and effective title for an article that you would consider perfect?
Output from GPT: Demystifying the Role of Prompt Engineers in the AI Era: A Universal Skill in the Making?
To make this work, you just need to activate the Prompt Perfect plugin before your prompt and use the keyword “perfect” at the end of the prompt. Admittedly, the plugin still needs some practice, but it’s just a question of time until it raises prompts to the level of a Prompt Engineer.
This makes the “How-Skill” one that can be universally applied and thus used by almost everyone. The main beneficiaries will be experts who master the “What-Skill” and are able to feed systems with the right context and also quality check the outcome. This practically makes everyone a Prompt Engineer, but will we still call it that?
Other interesting articles
It’s clear that AI will massively change the world as we know it; there will be winners and losers, but it won’t completely replace us humans (for now). Also, because there is a skill that makes humans irreplaceable (and it’s not empathy or creativity).
Sometimes it can also be helpful to reflect on the developments of AI from the perspective of philosophy. In this article, I asked six historical philosophers what they think about AI.
