Artificial intelligence (AI) is transforming the way knowledge is produced, research is conducted and students are taught. As machines become increasingly capable of performing tasks once reserved for highly trained specialists, universities face a fundamental question: What skills will graduates need when technology can do much of the work for them?
This dilemma was brought to the forefront during two forums at Stellenbosch University (SU) in October, where AI featured prominently in thought-provoking academic conversations about the future of higher education.
Although approaching the subject from different disciplines, two SU professors reached a strikingly similar conclusion: As AI simplifies the production of information and complex outputs, the ability to think independently, ask meaningful questions and exercise sound judgement becomes more valuable.
When producing knowledge becomes cheap
At an online Learning and Teaching Enhancement (LTE) seminar hosted by SU’s Centre for Teaching and Learning on 15 October, Prof Johan Fourie, professor of economics and Chair of Economics, History and Policy, challenged conventional assumptions about the future of academic work.
In his presentation, “AI and the future of the (South African) university”, Fourie argued that AI is fundamentally changing the economics of knowledge production. Tasks that previously required considerable time and specialist expertise, from analysing datasets to writing academic papers, can increasingly be performed by AI systems.
An enthusiastic advocate of the technology’s potential, Fourie described how he had used AI to reassess 30 of his own published research papers, identifying errors and weaknesses that conventional academic review processes had overlooked. For him, this demonstrated AI’s potential to strengthen scientific integrity by exposing mistakes, testing established findings and accelerating discovery.
However, the ease with which research can now be produced raises another challenge: If academics can generate hundreds of papers annually, who will read them, and how will worthwhile research be distinguished from an overwhelming volume of information?
Fourie suggested that academic attention, credibility and the ability to identify important research questions would become increasingly valuable. “Finding the question” would become central to the research process, he argued, requiring researchers to think more carefully about which problems deserve attention.
His observations extended to academic writing, where AI’s growing capabilities challenge the traditional relationship between writing and thinking. Fourie acknowledged the danger of students losing essential intellectual skills if they rely on machines before developing their own understanding.
He advocated using AI ambitiously while reconsidering how universities cultivate and assess independent judgement, including opportunities for students to demonstrate their understanding without technological assistance.
From writing code to knowing what matters
On 19 October Prof Willem Visser, professor of computer science at SU and Senior Principal Applied Scientist at Amazon Web Services, approached a similar dilemma from the perspective of software engineering.
During a colloquium hosted by the National Institute for Theoretical and Computational Sciences (NiTheCS) at the Neelsie Student Centre, Visser explored how generative AI is revolutionising computer programming.
His presentation, “English Is the New Programming Language (And We Are the New Bottleneck)”, described a world in which sophisticated software can increasingly be created simply by telling AI systems, in ordinary language, what they should do. “More software engineering, less coding,” was his prescription for the future of computer science education.
Drawing on his experience at Amazon Web Services, Visser explained how AI coding agents can generate enormous quantities of software, dramatically reducing the time required to develop applications. Yet producing more code does not necessarily mean producing better software.
The real challenge, he argued, lies in specifying precisely what a system should accomplish and verifying whether the generated output meets those requirements. Visser explained how automated reasoning, mathematical logic and rigorous software testing can help bridge the growing gap between the speed at which AI produces code and the ability to establish its correctness.
This approach places greater emphasis on understanding problems, designing reliable systems and evaluating results. Visser suggested that universities should devote more attention to these higher-level skills, preparing graduates to exercise the judgement traditionally acquired through years of professional experience. “Computer science is more important now than it used to be. It’s not less important,” he emphasised.
Teaching judgement in the age of machines
Perhaps the most revealing aspect of both discussions concerned an educational paradox: If students use AI to perform difficult intellectual tasks, how will they develop the expertise needed to judge whether its answers are correct?
Fourie acknowledged that outsourcing writing could undermine the learning that occurs when students struggle to formulate and refine their own arguments.
Visser encountered similar concerns during a lively question-and-answer session, when academics questioned whether students could develop problem-solving abilities without first learning to programme themselves. He conceded that there were no easy answers.
One approach he had experimented with was requiring students to submit their complete interactions with AI tools, allowing lecturers to evaluate the thinking and decisions behind the final product rather than judging the output alone.
Both discussions touched on broader challenges. Fourie considered unequal access to AI technology in South Africa, identifying internet connectivity and data costs as barriers despite his optimism about its democratising potential. Visser faced questions about the substantial electricity demands of AI infrastructure and the implications for countries with constrained energy supplies.
The perspectives of the two academics differed in emphasis. Fourie envisaged AI expanding the possibilities of scientific inquiry and urged universities to experiment boldly. Visser concentrated on the technical and educational changes needed to ensure that increasingly automated systems remain reliable.
Their shared message was nevertheless unmistakable: Universities must prepare students to work intelligently with AI while safeguarding the human capabilities that make knowledge meaningful. The future of higher education may depend increasingly on teaching students how to determine what is worth knowing, rather than simply how to produce an answer.
This report was compiled with the assistance of ChatGPT with human oversight by Hannelie Booyens.
©Higher Education Media Services.



