Africa is by no means “at the kiddies’ table” when it comes to contributing to engineering advancements in artificial intelligence, according to two professors who have made it their mission to support budding technology researchers on the continent.
When computer scientist Benjamin Rosman, who was “born and bred” in Johannesburg, was starting out in his career 20 years ago, most were forced to go abroad for postgraduate studies.
But Rosman, who co-directs the University of the Witwatersrand’s Robotics, Autonomous Intelligence and Learning Lab (RAIL Lab) with Steve James, says this picture has changed.
This is largely down to the efforts of Rosman and his collaborators to inspire machine learning knowledge development across South Africa and the wider region.
A central part of these efforts was the creation, nearly 10 years ago, of an initiative known as the Deep Learning Indaba, of which Rosman is a founder.
The event – named after the Zulu word for gathering – brings together scholars working in AI.
“A group of eight of us came together that were either South African expats or had studied abroad and came back. Some were researchers at Google and DeepMind,” Rosman said.
“We wanted to give something back to the community. We thought we’d do a deep dive workshop, a summer school basically, and we famously said, ‘if we really search far and wide and dig deep into our networks, we’ll find 30 people that care [about AI]’.”
The goal was to provide something for postgraduate students, researchers and academics looking to make progress on machine learning in Africa. But after expecting to host fewer than 50 people at an event first hosted at the University of the Witwatersrand, they ended up receiving about 750 applications.
Approximately 330 people joined that gathering, and the Deep Learning Indaba has grown to be the largest AI summer school in the world. Last year’s gathering, in Kigali, Rwanda, attracted 1,300 participants.
“It’s clearly the case that there is a massive appetite on the continent for AI research,” said James. “It’s definitely not the case that students are in limited supply.”
Those involved wanted to “strengthen African AI from something that didn’t [previously] exist as an entity”, Rosman said.
So rather than relying on one week-long annual summit, organisers set up a programme called IndabaX, which allows people from countries across Africa to run their own event exploring machine learning.
Successful applicants receive funding and help with speakers and advertising, with 48 countries out of Africa’s 54 hosting one of these grassroots events this year.
“This is now a network of tens of thousands of people across Africa that work together, that collaborate,” Rosman said.
As well as working to expand opportunities for talented researchers across Africa to get involved in AI, the pair are also both a part of Wits’ Machine Intelligence and Neural Discovery (MIND) Institute. Rosman is founding director of the institute, set up in 2024.
The indaba built “communication” across the continent, particularly between researchers. “Now we’re trying to build structures on top of that that allow us to capitalise on having that network and allow it to go deeper in different ways,” Rosman explained.
The MIND Institute studies intelligence, both artificial and natural, and involves some 65 academics, from behavioural psychologists to ethicists, neuroscientists and people looking at policy work.
The idea “isn’t new”, Rosman admitted, saying he liked to “benchmark” the institute’s work against that of places like Harvard University’s Kempner Institute for the Study of Natural and Artificial Intelligence.
But the MIND Institute leader suggested its work benefited from factors including proximity to the Cradle of Humankind, a Unesco World Heritage Site located about 50 kilometres north west of Johannesburg, which is home to the largest known concentration of human ancestral remains anywhere in the world.
“We’re in the leading place in the world looking at hominid evolution because there’s all these sites around the area including the Cradle of Humankind, where more hominid fossils have been discovered than anywhere else in the world, and that’s a research site of our university…This has given us a critical mass in very diverse ways of thinking.”
Rosman and James believe their work is proof African researchers are a crucial part of the equation when it comes to progressing AI and our understanding of variant forms of intelligence.
“A big push of ours at the beginning was, how do we make sure Africa’s at the table?” said Rosman. “Then I felt like Africa is at the table but we’re at the kiddies’ table. There’s this [attitude] to say, ‘oh, you’re doing AI in Africa, that’s so cool. You should take what we’ve built and use it to solve your problems’. That’s true and also incredibly condescending…My interest is in advancing AI itself.”
James, who was the first recipient of the Google Africa PhD Fellowship in the field of machine learning, added: “One of the risks we were trying to guard against is that Africa becomes a site for other people to do science.”
Instead, the pair believe they and their fellow researchers are building a community that is “asking interesting research questions and pushing the boundaries and frontiers of AI”.
“Africa has a billion people,” James pointed out. “If you are really passionate about robotics, like [for instance] you want humanoid robots in the world, why wouldn’t you want a continent of a billion people helping you with that?”
Leave a Reply