BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

The New ‘AI Psychologists’: The Rise Of Prompt Engineers

Following

David Lopez-Lopez is a Senior Lecturer at Esade Business School

The democratization of Artificial Intelligence and, specifically, the generative models boom seems to have changed everything. The same is true in terms of how we interact with machines. Conversational models such as ChatGPT or Bard and generative systems like Midjourney and Dall-e 2 are unpredictable and they are constantly learning. Consequently, to obtain quality answers, the questions we ask and how we ask them are increasingly crucial.

All this has led to the rise of a new job: the “prompt engineer.” This new professional is extremely well-remunerated and in high demand. On the one hand, these engineers are responsible for training AI with natural language and, on the other, thoroughly checking search results to create the perfect “prompts.”

Companies such as EY, NTT DATA, Datasumi and eDreams have already incorporated this position into their firms. According to LinkedIn Jobs and other sources, there are already more than 10,000 job offers that mention “prompt engineering” for different positions just in Spain, while approximately 100 are explicitly looking for a “prompt engineer.” Almost all of these job openings were published during the month of April. According to that same source and others, their corresponding salaries can range from approximately 95,000 euros per year in Spain to 300,000 in the United States.

The interest this new role has sparked is so intense, in fact, that the well-known expert Andrew Ng has launched a new course called ChatGPT Prompt Engineering for Developers in collaboration with OpenAI, the company responsible for the series of GPT and Dall-e generative models, among others.

Andrej Karpathy, another well-known scientist and co-founder of OpenAI, refers to these new professionals as “AI psychologists.” The idea behind this term is that psychology can play a crucial role in developing and applying this technology. Psychologists can provide insights on the human mind, cognition, behavior and interactions, something which may be fundamental to be able to design more effective, ethical and user-centric AI systems.

Our shifting relation with machines

The appearance of the “prompt engineer” concept reflects several interesting changes in terms of how we interact with machines. One is the way we communicate with them. This novel role adds a new chapter to an already extensive book that begins with the first human-machine interfaces based on “machine language” (only understood by expert engineers), followed by “front-end” engineers that encoded algorithms suited to human cognitive capacity and then by the idea of “usability” as defined by Jakob Nielsen. The novelty before the latest generative models was the even more ambitious CX or “client experience” concept which also incorporates people’s feelings before, during and after using technology.

Our ability to communicate by means of prompts has also changed the way we search for information. Thus far, internet searches have been based on how well we introduce the most appropriate keywords in search engines, marketplaces and social media to find the results we expect, results previously generated by other users for their own personal, social or professional use. Now, however, we have to ask better questions to find answers that haven’t been generated by others but, rather, have been self-generated by AI. In this new scenario, each question will receive a different, unique response; this explains why the way in which we ask our questions will become increasingly important.

These new “AI psychologists” have been defining strategies to obtain the best answers from machines. These strategies represent resources within any user’s reach and include, for example, clearly defining the roles played by users and the model in conversations, providing relevant contexts, defining restrictions in terms of content, tone and response length to guide the system and, lastly, segmenting the “prompt” into clear and defined steps.

Unwanted responses are also key for these professionals to make iterative adjustments and gradual changes until they obtain the desired results. That is why validating and revising responses is an important part of their work, as is experimenting with different focuses to improve the techniques’ interactions with the model.

A passing fad?

If “prompt engineers” have already become a reality in companies, some may wonder if they’re here to stay or if they’re just a passing fad. In either case, the existence of these new professionals can be framed within the human-machine interactions that have been underway for decades. First, the personal computer’s democratization prompted the mechanization and automation of processes and calculations, radically reducing times and costs. Later on, internet’s democratization changed the business paradigm thanks to the availability of free services in exchange for data used for commercial ends. As Andrew Lewis indicates in his well-known quote: “If you are not paying for it, you’re not the customer; you’re the product being sold.” Lastly, the smartphone’s democratization, mobility, and omnichannel experiences have brought about a profound change in how users behave and led to the “digital transformation” boom.

The question we now have to ask ourselves is: What will this new democratization bring about? Different theories abound, but perhaps the most interesting one, the one that encompasses the most concepts, is that AI’s democratization will serve to accelerate what authors such as Peter G. Kirchschläger and Marc Vidal refer to as the “Era of Humanity” and the “fifth industrial revolution.” “Prompt engineers” are just the first of the new jobs we’ll see arise in this new era.

Follow me on Twitter