Building a community to spread AI across Africa
But as AI investment booms around the world – it is estimated that AI will contribute $15.7 trillion to the global economy by 2030 – African experts are urging a more inclusive embrace of the technology, writes Forbes Insights
TFROM MEDICAL DIAGNOSES to translating the estimated 2,000 languages in use on the continent, artificial intelligence (AI) has been billed as a solution for a wide variety of challenges facing African countries. The issue is that AI solutions don’t spring fully formed—they emerge from communities of researchers and entrepreneurs, which require a lot of groundwork to build.
But with a growing cluster of research conferences, specialised educational opportunities and inventive forms of intra-continental collaboration, Africa’s tech leaders are stepping up—they’re building a core network of experts and students dedicated to using AI to help address some of the region’s many challenges.
In innovation hubs like Accra, Nairobi and Lagos, for example, technology companies have already opened AI centres to tap and foster regional talent. And major international organisations have argued that AI development in Africa is also a top priority given its potential to improve education, health, ecology and standards of living.
“How do we make sure Africans are actually shaping the innovations that are going to be used in this space?” asks Vukosi Marivate, chair of data science at the University of Pretoria in South Africa, where he leads the Data Science for Social Impact research group.
Marivate and others have been working hard to find answers to these questions. Here’s a look at some of the most promising efforts to spread AI across the continent.
Deep Learning Indaba: Building Local Networks
In 2017, in direct response to the inaccessibility of prestigious AI conferences, a small group of African data scientists (including Marivate) created their own.
Deep Learning Indaba, named after the Zulu word for gathering, aims to foster a community of AI researchers dedicated to the people and problems of Africa. Attendees range from deeply experienced AI engineers to graduate students and independent researchers. Each year, over a week of training sessions, seminars, research symposia and informal networking, they hone solutions designed to tackle local problems.
At last year’s conference, in Nairobi, the founder of a Tunisian AI company gave a seminar on how to build a startup. A Ph.D. candidate from the United Kingdom and a graduate of an African university hosted a session on AI and fairness with UNESCO.
The conference has been growing rapidly since the start, from 300 attendees in 2017 to over 600 in 2019. Many of the ideas hatched in its collaborative atmosphere live on. For example, contributors to one startup, Masakhane, are building machine learning models to translate English into many different African languages.
Doing so would address multiple problems. It would help English-speaking aid workers communicate more effectively with local communities. It would also give multiple countries access to relevant news and other content from around the globe— created in English, Arabic or French—translated accurately into African languages. The project, launched at Deep Learning Indaba 2018, now includes 70 researchers across the continent.
The ultimate goal of Masakhane’s venture is to submit a paper to a top natural language processing (NLP) conference, “and in doing so, once and for all put Africa on the NLP map.”
African Master’s In Machine Intelligence: Training The Next Generation
For African countries to realize the promises of AI—and other advanced technologies—they must build a faster and more accessible pipeline for those skills in higher education.
Case in point is Marivate, who grew up in Ga-Rankuwa, a settlement near Pretoria, but had to relocate to New Jersey to get his Ph.D. in computer science. Moustapha Cisse also had to leave the continent for graduate school, completing his Ph.D. in machine learning at a prominent university in France.
Cisse believed there could be a better way to address Africa’s talent shortage. In 2018, he launched the African Master’s in Machine
Intelligence (AMMI), based in Rwanda. As part of the African Institute for Mathematical Sciences, the one-year master’s program draws on a network of 1,682 alumni from 43 African countries to mentor and train the next generation of AI talent.
“The vast majority of active [machine intelligence] researchers and practitioners are in North America, Europe and Asia,” Cisse wrote in a letter introducing the program, “while large regions, including Africa, are hardly represented.”
Now in its third year, AMMI has become part of the broader continental community. Though not officially affiliated, AMMI and Deep Learning Indaba have been seeing crossover. For the master’s program, students work with professors at universities across the continent on capstone research projects. Indaba has seen a rise in the number of university faculty members attending, many of whom go on to advise the AMMI students.
That includes Marivate, who will be advising two students in this year’s program. He says this sort of collaboration strengthens the AI network across the continent, ensuring that both institutions will persist as places (both virtual and physical) of learning and idea-sharing.
Zindi: A New AI Skills Platform
Other initiatives help African organisations benefit from all this burgeoning AI talent. Zindi, for example, is an online platform that hosts open data-science competitions and hackathons for companies, non-profits and governments facing problems that could be solved with AI.
In March, the platform hosted three competitions aimed specifically at students. One, sponsored by the Ugandan start-up Xente, asked participants to create a machine learning model that could predict customer purchasing behaviour. Another asked for models that could predict when and where the most damaging wildfires would occur in the Democratic Republic of the Congo. Zindi announced in a tweet that the three competitions attracted students from 70 universities in 16 African countries.
Zindi has caught Marivate’s eye, too. One of the organizations he works with, AI for Development, is considering submitting a data set to Zindi for an NLP competition next year. With a project like Masakhane’s, Marivate says, posting a challenge on Zindi would further distribute its research across the continent. That’s the kind of promise for AI in Africa that every country can get behind.
Article first published in Forbes Insight