Who are prompt engineers and is it possible to master this specialty without knowing programming languages

In the field of artificial intelligence, a relatively young specialty has appeared – a prompt engineer, who helps to train an AI model so that it gives more accurate answers when asked.

With the advent of artificial intelligence models like GPT or Midjourney, it became clear that the bot is not always able to understand human requests. To solve this problem, a new approach has emerged in which you do not need to know a special programming language – you can communicate with an AI model in almost the same way as with a person. Prompt engineers own this specialization: developers who train artificial intelligence, find errors, generate answers using hints and commands.

How to get the desired result from the neural network

One of the main problems of artificial intelligence is that it does not always give accurate answers to human requests. Roughly speaking, ChatGPT is the same predictive dialing on a phone, only with mother-of-pearl buttons. To simplify a lot, they solve the same problem: guess the next word, based on the context, with the only difference being that language models like ChatGPT are trained on more information, and their algorithms are much more complicated than those in T9.

The task of a prompt engineer is to make the model increase the likelihood of giving the answers that are most truthful. This can be shown more clearly on AI models that provide images on demand.

Imagine a hypothetical situation: you are a graphic designer, and you have been given an urgent task to design some images for social networks in the style of the game Hogwarts Legacy. There is not enough time to draw from scratch, and here the Midjourney neural network comes to the rescue. According to the terms of reference in the picture, Neville Dogopups (or Longbottom?) was supposed to sit at the laptop screen. Here is what happened in the first approximation with a short text query:

This picture shows some guy with a macbook, which is hard to guess Neville Longbottom. Without context, it’s just a British man in a coffee shop, also with six fingers on his hand.

It is worth spreading the text query a little and the output is an acceptable result. Here it is already more or less clear that this is a Gryffindor student who is sitting in the library. Often, in order to produce a high-quality result, artificial intelligence needs an additional context that will decorate the picture – a kind of magic word.

*(Hogwarts student Neville Longbottom in Griffindor uniform, sitting in Hogwarts library with book shelf on background, looking at laptop on the table, confused face, realistic lightning)*

Second example. Again, an urgent task: to design the cover of the article, we need an illustration that depicts a man at a fork in the road, you can understand without words – everything is “perfect” here:

*(wide shot, man at the fork in the road, real photo, ultra realistic, photorealistic)*

The neural network took the word fork literally, and in the first request it drew two crossed forks, but in the second version, you only need to complete the request. The end result is a completely different result.

Let’s show an example of working with texts: the same request can look different if you add a small detail.

*So the profession of a programmer, according to AI, is seen by a gopnik and an official who fell ill with a clerk.*

An example with images shows that AI is not yet fully able to understand phraseological units, speech turns. Yes, it is often enough to clarify what exactly you need to get from the bot, but in this case, the requests look like a dialogue in the Literal Universe. Also, while most models are just learning: they do not know all the characters, professional terminology, facts from history.

Another problem of artificial intelligence that scares Internet users is that it may not always be aware of its violations. For example, AI models can produce images that are prohibited by law, disclose detailswhich can only be owned by developers.

What are the requirements for prompt engineers in the labor market

Now, when AI is not sufficiently trained, the profession of a prompt engineer is especially in demand: this area requires a lot of work. Easy entry into the profession is promised courseswhich claim – everything can be learned from scratch, you do not need any special knowledge.

On the Udemy website page, you can see that prompt engineering courses are among the top sellers, and new educational programs are now being released - the field is really young. At the same time, there is a risk of stumbling upon another “topic” – some businessmen sell ready-made prompts that can be found on the Internet for free. — *On the Udemy website page, you can see that prompt engineering courses are among the top sellers, and new educational programs are now being released – the field is really young*. *At the same time, there is a risk of stumbling upon another “topic” – some businessmen sell ready-made prompts that can be found on the Internet for free.*

We note right away that it is impossible to track the real number of vacancies among Russian employers: companies do not yet use the definition of “prompt engineer”. On job search sites, you can find a variety of wording – from “linguist developer” before AI trainers.

It should be noted that in the foreign labor market, not all employers use the clear wording “prompt engineer”. Here you can also find different variations: “prompt engineer and librarian”, “back-end developer with a focus on AI”, “engineer of AI-generated dialogs” and others.

Let’s consider what requirements Russian employers put forward for a prompt engineer using the example vacancies Yandex:

Ability to quickly understand various unfamiliar topics;
Text editing;
Skill in fact-checking;

Benefits for candidates include:

Experience with texts
Profile education in the field of linguistics, journalism or other related field;
Excellent editing and proofreading skills.

At the suggestion of Yandex, we can conclude that it is really possible to move from a humanitarian specialty. However, it should be understood that the amount of work, judging by the requirements, is really large: most likely, it will be handled primarily by data journalists or professional journalists who are not tied to the subject.

Consider foreign vacancy: Artificial Intelligence Prompt Engineer at Booz Allen. It offers a salary of 93.3 thousand to 212 thousand US dollars per year. Candidate requirements are as follows:

More than three years of experience in machine learning and other data processing methods in Python libraries;
More than two years of experience in creating statistical models and developing machine learning algorithms;
More than two years of experience in analyzing structured and unstructured data sources;
Over a year of experience in AI tooltip development using large language models including GPT-3 or ChatGPT;
Experience in natural language editing, including text mining.

As follows from this vacancy, the applicant will not only need the skill of working with text – it is necessary to have the proper experience in the field of artificial intelligence.

The Russian market is not rich in vacancies for a prompt engineer, and employers do not indicate at least a salary bracket, so it is difficult to estimate income for this profession. The spread in salaries in the foreign market is quite large: from less than 100 thousand US dollars to several hundred thousand dollars a year. On average, companies are willing to pay from $100,000 to $200,000. The fabulous money that bloggers talk about can only be held by AI professionals.

Can a humanist become a prompt engineer

Since prompt engineers do not use special programming languages in their work, the opinion is actively broadcast in social networks that anyone can become this developer in the field of AI. We add here the hype around artificial intelligence in general, and as a result, we will soon receive numerous low-quality courses that will offer to master the profession of a prompt engineer in three days.

You can often see that it is easiest for linguists, philologists, journalists and other representatives of the humanities to move into this area. Spoiler: far from it. I studied the requirements in the vacancies and checked if you need to have any Hard skills to become a successful prompt engineer.

Ex-CTO of the Technocracy Alexander Simonenko believes that loud statements that a person with experience in the humanitarian field can become a prompt engineer are real.

“I think it is quite possible for the humanities to move into this profession. A prompt engineer is a person who knows how to correctly select expressions that will be understood even with a play on words. Moreover, in most cases, a prompt is a set of theses of variables from one or two words. There is nothing difficult in this, it is more a matter of vocabulary and imagination.the expert explained.

At the same time, it is absolutely not necessary to have experience in the field of artificial intelligence: the one who had experience with a particular tool will look more advantageous. What gives a good result in Midjourney will not be so successful in DALL-E, the same with text-generating neural networks, even from version to version: the output from the same prompt for GPT-3 and GPT-5 will be noticeably different .

However, predictions that prompt engineers will take over the world of the future are unlikely to come true, according to the ex-CTO of the Technocracy.

“From the point of view of the demand for the profession, there is such a moment: what can be automated will be done in the future. For example, ChatGPT will be able to generate a prompt for Midjourney, and there will be little unique human creativity here. I think there will be a local outbreak of vacancies in the Russian and foreign markets within a year or two: it will appear quite quickly and then it will also start to decline,” Alexander concluded.

It can be concluded that in any case it will not work to enter the profession from scratch. It is logical that developers of various programming languages who move into this profession will have great advantages. For example, Scale AI employee Rally Goodside, one of the brightest representatives of the profession, has filled this niche with a knowledge base in machine learning, data science and, of course, artificial intelligence.

In addition, knowledge in the area in which the AI model works will be a big plus. If his task is to produce realistic photos on demand, then a professional photographer who is able to train the model will be a good prompt engineer. A copywriter who can complete courses on working with artificial intelligence in a short time will look better than a professional editor who does not have any knowledge of AI.

However, despite some uncertainty in terms of salaries, the lack of a clear understanding from some employers of what a prompt engineer should do, entering the profession at the very beginning of its inception in the future can lead to success. In a relatively short time, the sphere has a chance to become a desirable place for those who are interested in artificial intelligence.

Also subscribe to our telegram channel “Voice of Technocracy”. Every morning we publish a news digest from the IT world, and in the evenings we share interesting and useful articles.