Prompt:> Why does this conversation matter?

On 21 March 2023, Bill Gates declared that “the development of AI is as fundamental as the creation of the microprocessor, the personal computer, the Internet, and the mobile phone”.

Here, at The UJ Methods Lab, we concur.

However, a few days later, an open letter, signed by thousands of prominent technology figures and experts — including Elon Musk — called for “all AI labs to immediately pause for at least six months the training of AI systems more powerful than GPT-4”.

A day after that, Eliezer Yudkowsky, a decision theorist at the Machine Intelligence Research Institute, published an article in Time magazine, explaining that he had refrained from adding his signature to the letter because it was “understating the seriousness of the situation and asking for too little to solve it”.

Then, just a few days after that, the Italian data-protection authority announced that it would ban and investigate ChatGPT-4 “with immediate effect”.

Clearly, this is a “developing conversation”. But, the pace at which it is developing pales in comparison to the pace at which researchers in the field of AI have been continuously advancing and perfecting large language models (LLMs) — such as ChatGPT — which exhibit exceptional abilities across a wide range of domains and activities, pushing the boundaries of what we think we know about learning and cognition.

In a recent paper, researchers at Microsoft demonstrated that, besides its linguistic prowess, ChatGPT-4 is capable of tackling complex and unprecedented tasks in areas such as mathematics, programming, visual perception, healthcare, psychology and law.

Furthermore, they reveal that ChatGPT-4’s performance in all these domains is impressively (or, as some are cautioning, “dangerously”) close to that of humans.

In the business world, investors such as Bank Zero chairman, Michael Jordaan, are well aware that the introduction and adoption of ChatGPT signals that the “age of machine learning has arrived”, warning that businesses will have to “embrace AI or die” within five to 10 years.

In an interview with University World News, Pieter Geldenhuys – director of the Institute for Technology, Strategy and Innovation – puts it in a less apocalyptic way: “AI will not replace people — but the person using AI might replace you”.

Interestingly, he also suggests that AI will, “in a round-about way, reward original people”. This, he argues, is because generative AI models, like ChatGPT, do not produce anything “original”, but, rather, “repeat everything that everybody else is saying”, meaning that “humans need to become creative in using these tools”, especially in educational settings.

Similarly, the president of the Academy of Science of South Africa, Jonathan Jansen, argues that ChatGPT will “force” academics in the humanities and social sciences to become “much more inventive” in their pedagogical practice.

Prompt:> How can we use this technology more creatively?

On March 26 2023, Advaita Naidoo, MD of Africa’s largest executive recruitment agency, noted “a massive upswing in demand for prompt engineers in the past few weeks — a position that wasn’t even part of the mainstream lexicon two months ago”.

For a “job description” of prompt engineering, see The Washington Post’s excellent piece on “Tech’s hottest new job”. Essentially, an AI prompt engineer is responsible for developing and optimising prompts for AI language models. They work with data scientists, machine learning engineers, and other stakeholders to create high-quality, effective, and user-friendly AI interactions.

One need only conduct a literature search to appreciate the infancy of this vocation. On 24 March 2023, a Google Scholar search for “artificial intelligence” returned 3,160,000 results in 0.03 seconds. Adding “prompt engineer” to the search string generated just 76 results in 0.05 seconds.

Nevertheless, there are already hundreds of “prompt engineering” short courses available through online learning platforms such as Udemy, and there is even a book on how to enhance your interactions with ChatGPT through effective prompting techniques.

However, what most of these courses are referring to as “prompt engineering” is, essentially, the art of effectively interacting with generative AI models, so as to extract useful information through well-crafted questions and prompts.

Or, as Marcel Scharth puts it in his piece in The Conversation, “the skill of crafting an input to deliver a desired result from generative AI”. It is in this sense that we deploy the term in the present article.

In this context, being a good “prompt engineer” would mean being skilled at, for example: formulating clear and concise prompts to elicit relevant and informative responses; asking thought-provoking questions that encourage deeper analysis and more nuanced answers; and iteratively adapting prompts, based on the AI’s responses, to guide the interaction in a specific direction or obtain more detailed information.

In this sense, prompting can be thought of as the “language” required for human-AI interaction — a “lingua franca” that facilitates collaboration and communication between humans and machines. We believe this lingua franca holds great potential for enhancing human-human interactions, too.

Prompt:> How could AI’s lingua franca prompt progress in Africa?

By leveraging AI’s capabilities, prompt engineering — both the profession and the skills we have described — has the potential to unlock creative solutions to Africa’s most pressing challenges, across domains such as healthcare, agriculture, and renewable energy. Here are a few ways in which we imagine this could happen:

1. Fostering interdisciplinary research: as a shared language, AI’s lingua franca can bridge the gap between disciplines, granting researchers access to knowledge outside their fields, promoting collaboration;
2. Expediting data analysis: effective prompt engineering can optimise AI interactions to help researchers sift through datasets, write code, identify patterns, and extract valuable insights, benefitting fields like epidemiology, climate change, and the humanities and social sciences. The UJ Methods Lab has been interacting with ChatGPT-4 for qualitative analysis, with astonishingly robust insights;
3. Supporting local languages and cultural contexts: Africa’s diverse languages and cultures can be catered to by developing AIs that communicate effectively across geopolitical and sociocultural boundaries. This democratises access to technology, drives adoption, and addresses regional challenges;
4. Fostering innovation in education: AI is reshaping educational practices, facilitating personalised learning experiences, helping educators design more engaging critical pedagogical resources that promote thinking outside the classroom. As Pete Rorabaugh writes: “If students live in a culture that digitises and educates them through a screen, they require an education that empowers them in that sphere, teaches them that language, and offers new opportunities of human connectivity”;
5. Boosting economic development: with the growing demand for prompt engineers, Africa can develop a skilled workforce of tech-savvy innovators to create job opportunities, attract global investments, and drive economic growth; and
6. Addressing social challenges: AI-driven education, entrepreneurship and application can help to break the back of poverty, inequality, environmental degradation and other plagues that have kept Africa in bondage.

Prompt:> Start prompting, promptly (but carefully)

In his article in Time, Eliezer Yudkowsky warns that if we continue to allow AI models to evolve at their current pace, “everyone will die, including children who did not choose this and did not do anything wrong”.

Yet, millions of children — children, who have not done anything wrong, and who would choose life over death, if the choice was theirs — are dying, anyway.

According to Unicef, in 2021, one in 14 children in sub-Saharan Africa died before reaching their fifth birthday, largely from preventable causes. It is abundantly clear that AI has the potential to change this — for better or for worse. By approaching prompt engineering as a lingua franca for human-machine interaction, we are prompted to consider its implications for human-human interactions, too.

We are prompted to think through how (open) access to this language is intertwined with the democratisation and decolonisation of knowledge.

It is a prime time for researchers to consider how this lingua franca could be leveraged to tackle Africa’s most pressing problems, foster economic growth, and secure a brighter future for our people.

Becoming fluent in the lingua franca of AI is, suddenly, about infinitely more than simply engineering good prompts.

It’s about engineering our future.

In the words of Bill Gates, “It’s hard to imagine a better use of AIs than saving the lives of children”. However, it will be up to us to prompt them to do so. DM

Dr Clinton Rautenbach is a postdoctoral research fellow at the University of Johannesburg’s UJ Methods Lab, promoting open access to open science in Africa. Professor Kevin Durrheim is Distinguished Professor in Psychology at the University of Johannesburg, where he heads the UJ Methods Lab. He is an NRF A-rated scientist.