Manamela sets universities a double AI task, and only half of it is theirs to control

Universities have been handed a two-part brief on artificial intelligence, and only one part of it sits fully inside their own gates.

The instruction came this week from the minister responsible for the sector, who used the 4th Universities South Africa conference to set out what he expects from the country’s campuses. Graduates, he said, must be able to work with AI tools capably and to interrogate them with a critical eye. Universities must also turn out researchers who can contribute to building the technology itself.

The first half of that task is largely a matter of curriculum, teaching and assessment, and universities control it. The second half is a different proposition. Producing people who can construct AI systems depends on policy settings, on access to data and on computing power, and the country is still in the early stages of putting those pieces in place.

That gap between what a university can teach and what a researcher needs in order to build is the substance of the test Manamela described. A graduate who knows how to use a tool well can be trained in a lecture hall. A researcher who can help design the tool needs more than a syllabus. The work requires machines to run models on, datasets to train and test them against, and a policy environment that tells institutions what they may do with both.

For students, the practical consequence is straightforward. The qualification they leave with will increasingly be judged on whether they can operate in a workplace where AI tools are already present, and on whether they can tell when those tools are wrong. That is a skill question, and it is one universities can answer with teaching.

For the research side, the consequence is slower and less certain. Building AI capability takes sustained investment in computing infrastructure and in the people who maintain it. It also takes a supply of data that researchers can lawfully use. Where those are absent, talented graduates tend to go where the infrastructure is, and the country loses the capacity it was trying to create.

The conference setting matters. Universities South Africa is the body that speaks for the sector as a whole, and the 4th gathering of its kind put the question of AI directly in front of vice-chancellors, academics and administrators at the same time. The message was not aimed at a single institution. It was aimed at a system.

Manamela’s framing drew a line between using and building, and that line is where the policy work now sits. A country that only consumes AI tools is dependent on decisions made elsewhere, on pricing set elsewhere and on priorities set elsewhere. A country that builds them holds more of its own future. The minister’s point was that universities are the place where the shift from one to the other has to begin.

What the sector does with the instruction will vary. Some universities already run programmes in data science, machine learning and related fields. Others are further back. The conference gave the sector a shared statement of intent. It did not, on its own, supply the computing power, the data access or the funding that the second half of the task requires.

That is the honest position. The first job is a teaching problem and universities can get on with it. The second is a national capability problem, and it will be solved or not solved by decisions taken well beyond any campus.

Source: IOL, From Users to Builders: The Test Facing South Africa’s Universities