The dawn of the AI Prompter
By Martin Weis, Managing Partner and EMEA Head of AI at Infosys Consulting
The launch of AI chatbot ChatGPT has sparked both enthusiasm and fear. How will the world of business change? At the same time, a new job title came into the public’s focus: AI Prompter, or Prompt Engineer.
The ability to compose prompts for AI tools like ChatGPT or the image generator DALL-E in a way that generates optimal results is now a sought-after skill. But what does this role look like in reality?
The power of prompting
Prompt engineering is a lot like a Google search: the more accurate the input, the better the output. ChatGPT as a language processing tool was trained with vast amounts of data from the internet and books, and can answer questions, generate stories, write software code and generate text from data. The prompt in large language models (LLMs) like ChatGPT can range from a simple question to a complex topic with a multitude of facts. Precision is key when it comes to optimising prompts.
An AI prompter understands how to communicate with the AI in a way that allows it to reach its full potential. Successful prompts combine instructions, questions, data and examples. Better results can be achieved by asking the AI to assume a certain identity – for example, a text written from a journalist’s perspective will be different to one written from the perspective of a scientist. Prompting generates high-quality training data that enables the AI model to make accurate predictions and decisions. This is an essential step in the evolution of AI systems.
Despite all these prompting skills, it is always important to take all generated results with a grain of salt. Biases – stemming from training data – can creep into the generated responses and AI models like ChatGPT can also hallucinate, providing wrong or illogical answers.
What does it take to become a prompter?
There is no exact profile for AI prompters. While programming skills are not mandatory, they are helpful in systematically evaluating prompts. Basic knowledge in machine learning makes it easier to understand the strengths and weaknesses of AI models and to adapt the prompts accordingly. However, there are some successful AI prompters with a humanities background too. Knowledge about human thinking, behavior, and psychology can be beneficial, especially to help develop ethical and user-friendly AI systems.
To have success as a prompt engineer, it is essential to have a very good understanding of language to be able to formulate the prompts precisely and effectively. During the “conversation” with the AI system, prompts can be refined.
Analytical thinking is important in prompting to help identify patterns for solutions, while critical thinking enables the evaluation of results. Prompters should be curious people who like to solve puzzles – persistent, but also flexible.
Are the best AI prompters people who are highly competent in both coding and linguistics? Some experts see this (rare) combination as a sought-after super talent. Others are convinced that older people can prompt better because of their greater knowledge and wealth of experience. What profile companies are looking for also depends on the employer’s field – so it’s advisable to have specific industry knowledge.
Prompting across multiple professions
Only a few offers for prompt engineering can be found on job portals. But we can expect these to increase as understanding of generative AI’s ability to increase productivity for companies widens. Marketing and advertising agencies are among the first companies to hire AI prompters, as there are high hopes for AI systems like ChatGPT in these sectors, especially when it comes to content creation. The fields of application for AI prompters can vary widely. From software developers using generative AI to write and check code, to those using it to develop chatbots for customer service, extensions for search engines, or optimising the accuracy of recommendation algorithms. Use cases for generative AI cover all industries and include content creation and communication, business performance reporting and risk management, as well as predictive maintenance in manufacturing or medical diagnostics.
There already is a wide range of courses and trainings on AI prompting – from YouTube videos to LinkedIn Learning webinars and courses offered by popular training portals such as Coursera or Udemy. Those who have the recommended skills and want to follow this career path will be spending a lot of time experimenting with generative AI at first. In the meantime, there are numerous plug-ins for large language models such as ChatGPT that optimise prompting, e.g. PromptPerfect. Interested parties who tackle the topic now will benefit from the first-mover effect.
Is prompting here to stay?
Will the profession of AI prompter establish itself in the long term? Currently, we can only guess. Many AI experts believe that, in the medium-term, prompting will become a core competency expected of all employees in many areas – similar to what happened with Microsoft Office. AI programs will become continuously better at recognising what users require and generating high quality responses. The ultimate goal is not to interact with a chatbot by entering long and complicated prompts, but to simply “talk” to it.
A growing number of companies will train their own large language models and employees will become more skilled at using generative AI. In general, it is advisable for companies to train their employees and familiarise themselves with these new tools. At the same time, management should develop an AI guideline for the correct handling of AI, especially regarding data protection and transparency, because it can be assumed that in the future the increase in productivity gained through generative AI will become a competitive advantage.