Universities must re-model themselves around Artificial Intelligence

Universities may cease to be as relevant as they are today if they don’t change their way of teaching and their curricula in the face of the Artificial Intelligence (AI) revolution.

This was the view of Professor Ulrich Paquet, Director of the African Institute of Mathematical Sciences (AIMS) South Africa, expressed during a plenary sub-themed  Technology and the Human Interface in the Future University at the 3^rd Higher Education Conference of Universities South Africa. The Conference was themed “The Future of the University”.

“We see the effects of disruptive innovation on companies like Kodak – once the leader in film and consumer cameras – and Nokia who once maintained a hold over the global mobile phone market.

“Universities need to consider that their business models can be disrupted – we cannot do the same business as usual and  continue to operate on a horizontal line. Universities need to operate in a space where we think about vertical or intensive progress and start doing new things. If you don’t keep an eye on both axes, then you are often out innovated,” he said.

He believes the business model of universities, in its present form, might be obsolete in two decades for the following reasons:

• All  the resources and the course material that students need are available for free online – this applies to first year students at AIMS and second and third year students in Stellenbosch unless you are studying medical science or a related field.

• There is a proliferation of Online Open Courses, and
• Free and open courses do not put students into massive debt when they graduate– a prevalent burden today.
• All the components are therefore on the table for the university system to be radically disrupted if leadership is not careful. 

This was not the first time that scholars were debating about the future of the university’s model. Professor Paquet referenced English writer Samuel Johnson who speculated in as early as 1781 that “Lectures were once useful; but now when all can read, and books are so numerous, lectures are unnecessary”. He added that universities would remain relevant because of the learning climate that they provide.

Professor Paquet went on to say he had learnt from discussions with students that they desire to be integrated into a social ecosystem which none of the online learning experiences can provide. Furthermore, students benefit from extrinsic motivation, or even peer pressure, to lead them through a curriculum.

When they first made their appearance, massive open online courses (MOOCs) promised three things – expanded access, transformed systems and better teaching from data driven resources. However, Professor Paquet does not believe that they are all they initially promised to be.

“Successful MOOC learners demonstrate effective use of self-regulated learning strategies such as goal setting, time management, help seeking and self-monitoring – these are often the invisible things and the invisible curriculum that’s taught at the university,” he explained.

“Eighty percent of people who actually complete one of these MOOC courses have a bachelor’s degree already; so, what you find is that all of these technologies serve the educated and those pursuing post-secondary education and this, in turn, creates a market cycle. Online courses drift towards education for people already in business or who want to upskill themselves…”

However, Professor Parquet believes that the components are still on the table for some company or institution to disrupt education and make these courses cheap and affordable.

He said it is AI that is forcing universities to rethink how they do things and how they teach and examine in the future.

“A modern AI system allows a student who is tech savvy to write a prompt which will give that person input from a stack of history text books within 30 seconds; out comes an entire essay written as a summary of the text books,” he said.

But how do we see the wood from the trees and understand the core of the AI revolution?

Professor Paquet rapidly displayed a number of images (Mickey Mouse with President Ronald Reagan, Nelson Mandela with Francois Pienaar among others) and then proceeded to ask a number of questions about what the audience had seen.

“You may associate the picture of South Africa winning the 1995 Rugby World Cup with a memory of a braai because that is what you did on that day all those years ago. This demonstrates the kind of work that US physicist John Hopfield, who recently won The Nobel Prize in Physics for 2024, has undertaken. It is called a Hopfield network (or associative memory) which is a form of recurrent neural network.”

Professor Paquet posed another question: “What is 0.4235 X 1.84?” which nobody could answer in a matter of seconds except with a calculator.

“AI is nothing but taking the complicated questions that you’ve just seen answered and doing exactly the same, but on the computational substrate that’s nothing more than a traditional pocket calculator. The pocket calculator and the computer only see things as numbers. The computer doesn’t see Mickey Mouse, it only sees a long series of numbers. It doesn’t see history textbooks; it sees one very long series of numbers. If you zoom into a picture, each pixel is represented by a number.”

None of these new innovations have been in the market for more than a decade. AI’s accuracy in understanding speech and language is growing better and better, faster and faster.

These models or artificial brains have lots of numbers in them, typically have more than a 100 billion parameters. These parameters can be “tuned” because of a combination of factors including progress in computation, progress in “model architectures”, progress in algorithms and increased data set sizes.

AI has even blindsided the experts, said Professor Paquet: “As these artificial brains grew, up until about two years ago everyone could predict where technology was going. But at a certain scale, emergent behaviour was unlocked which wasn’t predictable, enabling AI to do many tasks that we originally didn’t think were possible.”

This includes question answering, semantic parsing, proverbs, code completion, general knowledge, reading comprehension, summarisation, logical inference chains, common-sense reasoning, pattern recognition, translation, dialogue, joke explanations, physics Q&As, language understanding, law and medical examinations.

The biggest disruption to education comes as AI answers complex mathematical equations and writes essays. 

“This means that we have to physically interview every person who applied to us at AIMS to know whether they actually submitted legitimate work. It changes the way we think about teaching. It changes the way we think about examinations and  the way students are assessed. I believe curriculums should be redesigned from scratch.”

“In fact”, he said, “we should run open competitions to design new curricula!”

With reference to the former President of the European Research Council (ERC), Helga Nowotny, who addressed audiences regarding the control of AI in Cape Town this year, Professor Paquet said: “The way educators think about assessing students in universities has to change completely. There’s no way we can go back to essays or any submitted assignments. 

“I asked her how she would do this? And she said, ‘run an open competition with different teams to submit curricula and evaluate them. Don’t even try to do it on your own’. Although it will be disruptive, it is a very exciting future. My hope for all of us here, in Africa, is that we will be ready and equipped to embrace it,” he concluded.

Janine Greenleaf Walker  is a contract writer for Universities South Africa.