‘AI Does Not Eliminate The Need For Human Judgment’

Professor Tshilidzi Marwala’s earliest lesson on innovation began in Duthuni village in the Tshivhasa region of Venda in South Africa’s Limpopo Province. He watched his grandmother, a master potter and mat weaver, deftly tap her clay pots and listen carefully to the sound that told her they were structurally sound. 

“She was, in every sense, except the credentials, an engineer, and she taught me about materials, precision, and design before I ever set foot in a classroom,” says the African professor, who is also an author.  

As one of the leading voices in artificial intelligence on the continent, he is also currently the seventh Rector of the United Nations (UN) University and a UN Under-Secretary-General.   

But what attracted him to AI specifically?  

Marwala says it was neural networks and the challenge of creating machines that must make decisions under uncertain circumstances, which sits at the intersection of engineering, economics, and philosophy.  

“My research has focused on developing rational machines while ensuring that human oversight remains essential, since AI does not eliminate the need for human judgment,” he adds. 

He also believes that the role Africa should play in the global ecosystem should not be as passive consumers of models built elsewhere.  

“That is tantamount to cyber-colonization, where the continent again ends up a rule-taker in a revolution it didn’t help design.”  

Marwala further explains that the continent has real assets such as demographic weight, its own compute and data initiatives, the African Union’s Continental AI Strategy, and diplomatic gravitas with bodies like the UN.  

The continent’s role should be threefold, he says. The first is that it needs to start with investing in African-owned data and compute infrastructure so value isn’t entirely extracted. Then, the continent should build AI talent and research capacity at scale through universities.  

And the final step involves showing up substantively in global governance forums rather than governance being negotiated, primarily, among a handful of AI-producing states. 

“What excites me is the possibility that, if we act deliberately, AI can narrow rather than widen global inequality,” he enthuses.