As soon as a new technological innovation comes along, there are a number of news articles about which jobs will disappear because they will be replaced by this new technology. This is, of course, a logical reaction to a new scenario. This is just a situation that has been experienced before. There will be certain jobs that may disappear, but new ones will appear linked to these innovations, as is the case with Artificial Intelligence, AI. So, will prompt engineering be the job of the moment?
What does prompt engineering consist of?
The concept of prompt refers to the set of prompts, using PLN natural language processing, to create the linguistic model that is fed into an AI application, with the aim of obtaining a relevant result that answers the query. Meanwhile, prompt engineering is the discipline that deals with refining the language models used by AI tools. In this way, through the design and infinite testing of inputs, phrases and questions, these specialists make a system provide the most suitable answers required by the user.
It is, therefore, a specific area of data science which, according to estimates published in Trend Statitics, will grow by 36% in the coming years.
What is the role of the prompts engineer?
One of the particularities of this speciality is that it is not a typical engineer with technical knowledge. In fact, he or she does not need to know how to programme because he or she will not be entering code or developing any software applications. Instead, he or she must have communication skills and knowledge to ensure that an AI-based system can provide accurate responses.
These are some of the main tasks in prompt engineering:
- Interact with the system to test it. Through the use of different grammatical structures and commands, this professional evaluates the system’s ability to answer specific questions. If he/she detects that, depending on the way of asking, the system varies its answers, he/she starts training it.
- System training. The language models, based on natural language, used by AI need to undergo training to adapt to the needs and demands of each user. It is a technique used to optimise the instructions or commands given to the language models in order to improve the quality of the responses generated. It is not an application trainer per se, but an approach to improve the performance of AI-based applications. That is, it aims to improve the quality and relevance of the generated answers. Engineers experiment with different formulations, formatting adjustments and adding context to get more accurate and useful results.
- Modulate the tone. During the AI testing and training process, this professional can modulate a certain type of tone for a certain type of user. For example, he can set a playful tone or a formal tone, thus offering personalised tools.
- He contributes to the optimisation of the programme. It works hand in hand with the IT and IT development team. This is because this is the person who detects if there are any errors in the database that feeds the technology or if further training is needed before launch.
- Delineate the functions of the models. Prompt engineers make sure that the tool does its job. If it’s a chatbot to provide information about a hardware store’s products, it won’t talk about health issues, for example.
What skills does a prompts engineer need to acquire?
You need to have a combination of technical and creative skills to carry out your job effectively. The most important of these are:
- Language model knowledge: This involves knowing the technical details of the specific model you are working on, its architecture and capabilities.
- Understanding the task and the context: It is essential to understand the specific task or problem for which the prompt is being optimised. You must also have a solid understanding of the domain and context in which the language model will be applied.
- Creativity and lateral thinking: You need to think creatively and find new ways of posing questions or instructions to get more precise and relevant answers. You need to be able to devise different approaches and experiment with them.
- Analysis and evaluation skills: You need to have the right tools to analyse the results generated by the model and evaluate the quality of the answers.
- Linguistic knowledge: You need to be proficient in grammar, syntax and semantics to formulate commands and questions effectively. In addition, he/she must have the ability to understand how words and linguistic structures affect the model’s responses. To some extent, you should be a philologist with a technical background.
- Basic programming experience: You should be able to manipulate and adapt commands as needed.
- Ability to collaborate and provide feedback: Typically, you will work as part of multidisciplinary teams.
Although prompt engineering is a newly created profession, it is already in strong demand and will grow in the coming years. It requires a combination of skills hitherto unusual in our labour market, and its emergence is an example of how work is evolving in today’s societies.