Researchers want to pave their own path. But the growing industry is still dependent on tech giants like Google and Microsoft.
In late August, under the shade of an arching pepper tree in Nairobi, Kenya, hundreds of A.I. researchers gossiped about their algorithms. Some stood in front of posters, which wound around the tree’s sprawling roots, depicting machine learning systems that promised to predict everything from soil nutrition, to whether a small-scale farmer would repay a loan, to how a self-driving car might navigate the bustling streets of Cairo.
Over the last three years, academics and industry researchers from around the African continent have begun sketching the future of their own A.I. industry at a conference called Deep Learning Indaba. The conference brings together hundreds of researchers from more than 40 African countries to present their work, and discuss everything from natural language processing to A.I. ethics.
Founded in 2017, Indaba is a direct response to Western academic conferences, which are often difficult for researchers from distant parts of the world to access. Take, for instance, the Conference on Neural Information Processing Systems, the most well-known meeting dedicated to artificial neural networks. NeurIPS — originally referred to as NIPS until the community overwhelmingly asked for a less nipple-oriented acronym — has previously been held in distant and expensive resorts. It doubles as a kind of vacation for researchers who can afford it. In 2006 and 2007, it was at the Westin Resort and Spa, and Hilton Resort and Spa in Whistler, B.C, to allow for “informal discussions, skiing, and other winter sports.”
For researchers from Africa, NeurIPS is often just out of reach. In 2016, no papers from African countries were accepted into the conference. In 2018, more than 100 researchers were denied visas to enter Canada for NeurIPS.
We need to find a way to build African machine learning in our image.
So in 2017, former classmates from South Africa’s University of Witwatersrand and a few close colleagues came together to found Indaba, which they named after a Zulu word meaning “an important conference or gathering.”
“How many accepted papers have at least one of its authors from a research institution in Africa? The answer: zero,” Indaba organizers wrote in a blog post. “Two entire continents are missing from the contemporary machine learning landscape.”
Organizers expected about 50 people to come to the first Indaba, but nearly 750 applied, and 300 were invited to attend. In its second year, Indaba invited 400 people and expanded into 13 IndabaX conferences. This year the conference nearly doubled in size again, with 700 attendees and 27 IndabaX events.
Deep Learning Indaba has become connective tissue for the African A.I. community — not only the space for the community to meet, but a part of the community itself. The conference forges relationships between researchers on the continent with a clear agenda: to build a vibrant, pan-African tech community — not through reinventing existing technologies, but by creating solutions tailored to the challenges facing the region: sprawling traffic, insurance claim payments, and drought patterns.
Google, Microsoft, Amazon, and other tech companies underwrite Indaba to a ballpark figure of $300,000, but organizers are still adamant about creating a new, distinct field of research — an industry free from the grip of Silicon Valley.
As Vukosi Marivate, an Indaba organizer and chair of data science at the University of Pretoria in South Africa, told me, “We need to find a way to build African machine learning in our image.”
This year’s Deep Learning Indaba was held over the course of six days in Kenyatta University, which sits just off Nairobi’s Thika Road, a bustling 8-lane highway full of boda boda motorcycle taxis and busses shuttling crowds to and from the city center.
Students make up a large portion of Indaba attendees — a chief reason why the conference focuses so intensely on education. The first day of the conference was dedicated to A.I. refreshers and introductory courses, such as statistics and the basics of building neural networks. Over the course of the week, the courses ramped up to more advanced topics. Attendees went to specialized courses on natural language processing, computer vision, deep reinforcement learning, and ethics. Some participated in a hackathon, where they built A.I. that could automatically identify African wildlife to better study and protect endangered species, while others worked with health data to predict and control the spread of malaria.
Establishing a presence in Africa now means building valuable relationships with users from the very start of their digital lives
Days typically started with a keynote: IBM researcher Aisha Walcott-Bryant spoke about data gathering in partnership with the Kenyan government; Princeton University’s Ruha Benjamin gave a course on the inequity encoded into algorithmic systems; and Salesforce’s chief scientist Richard Socher spoke about his team’s research on building more general A.I. systems.
As the preeminent A.I. conference in Africa, Indaba attracts significant attention from tech’s biggest players. American companies, including Google, Microsoft, Amazon, Apple, and Netflix, make up 11 of Indaba’s 34 sponsors.
It’s part of a trend in Silicon Valley firms making significant investments across the continent. Google sponsors organizations like Data Science Africa and the African Institute of Mathematical Sciences. In 2018, the company announced its first African research center in Accra, Ghana. Meanwhile, Microsoft and the Bill Gates Foundation have donated nearly $100,000 to Data Science Nigeria, which hopes to train one million Nigerian engineers in the next 10 years.
The appeal of investing in Africa is obvious. 75% of the continent still has no internet access. That’s a challenge for local populations, but also an investment opportunity for international tech firms. Establishing a presence in Africa now means building valuable relationships with users from the very start of their digital lives. A recent investor report indicates Facebook stands to make $2.13 for every user per year in the developing world — up from $.90 in 2015.
Chinese firms have spent years investing heavily in Africa’s tech infrastructure. Huawei installed surveillance cameras around Nairobi on behalf of the Kenyan government, and the company is currently working on large-scale facial recognition surveillance systems around Zimbabwe.
All this foreign investment and data collection raises red flags on a continent scarred by exploitation. Observers like PhD candidate Abeba Birhane argue that these efforts recall previous colonization efforts. “This discourse of ‘mining’ people for data is reminiscent of the colonizer attitude that declares humans as raw material free for the taking,” she wrote in a recent paper titled The Algorithmic Colonization of Africa.
Indaba organizers are well aware of the tension inherent in running an Africa-focused conference funded by American companies.
“A large part of the Indaba’s funding came from international organizations, and many of our international speakers came from international technology companies,” founders wrote after the first conference. “This risks leaving the impression that the best work is happening in the large tech companies, and in countries outside the continent, and that one must leave the continent to have an impactful career in the field.”
That’s why organizers try to balance international sponsors with local ones, and highlight opportunities at African universities and tech companies.
This year’s Indaba conference presented the Maathai Impact award to Bayo Adekanmbi, chief transformation officer of South African telecom MTN, and the founder of Data Science Nigeria (DSN), an organization that has trained tens of thousands of Nigerians to bolster the country’s IT sector.
“A large part of the Indaba’s funding came from international organizations, and many of our international speakers came from international technology companies.”
Adekanmbi arrived at the conference wearing a crisp, white short-sleeve button-up shirt tucked into jeans. He’s tall and broad and walks with a briefcase in one hand, looking ready to evangelize the gospel of data science. He’s revered by his students.
Adekanmbi wants to train a million Nigerian data scientists in the next 10 years, and as such, he takes the threat of a brain drain very seriously. “It’s a big concern. Talent always moves to the area of highest concentration,” he says. “If we do not build another community here, talent will continue to disperse.”
One way of keeping talent at home is remote work. Adekanmbi started a program called Data Scientists on Demand, which facilitates engineers in Nigeria to work remotely for companies around the world.
But that’s not enough. Adekanmbi says that corporations that want to reap the benefits of doing work with Africa should have a physical presence on the continent, and show commitment by investing in local tech communities.
“If you really want inclusion, fairness, and the distribution of knowledge for local relevance, then corporations should be willing to set up centers of excellence and create centers of knowledge around the world,” he said.
Karim Beguir, cofounder of Tunisian A.I. company InstaDeep, built his company after moving back to Africa from a finance career in London. It’s the prototypical startup story: leaving a cushy career with dreams of starting a new company with only two people and two laptops. He taught one of the introductory mathematics courses at Indaba, and ran a two-session seminar on building a start-up: how to find a cofounder, how to get the best tax advantages, and the basics of courting investors. Beguir says he isn’t worried about an African brain drain — he sees own trajectory as a model for cross-continental collaboration.
For many in Africa, partnerships with international tech giants are critical to conducting their work. Indaba attendee Tejumade Afonja, an engineer from Nigeria who cofounded a 16-week A.I. coding workshop affiliated with the global A.I. Saturdays organization, says her Intel sponsorship keeps the entire effort afloat. Intel asks that the organizers and instructors give students the option of using Intel software and hardware, but doesn’t make demands in return for the money, Afonja says.
Teki Akuetteh Falconer, former executive director of the Ghanaian Data Protection Commission and founder of Africa Digital Rights Hub, says that U.S. tech companies are some of the only organizations with aligned interest in expanding internet infrastructure and studying technological ecosystems. “You honestly can’t run away from it. I’m running an NGO, and I have to fund it,” Falconer says. “My resources cannot reach very far. And the strange thing is that the only ones who get me are these [internet] companies.”
Deep Learning Indaba is still looking to expand. A phrase that’s heard over and over from those interested in building the future of African artificial intelligence is “capacity.” To bring more people into the field on the continent means a greater legacy of researchers, with the capacity to teach the next generation. It’s also about Deep Learning Indaba as a name in the global deep learning community.
Part of that process is creating Indaba chapters in individual countries. Between countries like South Africa, Senegal, and Somalia, there are now 27 IndabaX events, which borrow the “X” terminology from the TED conferences.They range in size from just a few people to dozens, with competitions for “effort and excellence,” according to the competition outline.
Competition winners and IndabaX organizers are then invited to the main Indaba conference. Once there, two winners of the Deep Learning Indaba continent-wide poster session get sponsored to go to NeurIPS, the top industry event. The goal is to eventually place A.I. researchers in Africa on equal footing as those in the West.
“Right now we’re telegraphing the movement. Now you’ll see at NeurIPS an [Indaba] delegation that comes that is there on an equal footing,” says Marivate. “They are not getting there on ‘Oh, these Africans, we should bring them,’ but at the table as an equal.”
Update: A previous version of this article misidentified the cofounder of InstaDeep. His last name is Beguir.