Africa is at a crossroads in its journey to establish itself as a global force in artificial intelligence, but obstacles such as limited funding, infrastructure challenges, and fragmented regulation have left the future of the sector uncertain. As AI gains momentum worldwide, African researchers, innovators, and entrepreneurs are increasingly determined to carve out a space for AI solutions that address the unique needs and opportunities of the continent.
This determination was palpable at the recent Deep Learning Indaba, an annual AI conference held this September at Amadou Mahtar Mbow University in Dakar, Senegal. The event brought together over 700 attendees from across Africa, showcasing cutting-edge research and providing a platform to discuss the transformative potential of AI across sectors like healthcare, agriculture, and education.
A Young Vision for AI Innovation
Among the attendees was Kessel Okinga-Koumu, a 28-year-old computer science student from the University of the Western Cape in South Africa. Okinga-Koumu presented a prototype app she had developed, which uses virtual reality and AI to simulate lab equipment, allowing students at her university to engage in realistic, hands-on lab experiences despite a shortage of physical resources. Her app exemplifies the ingenuity of African researchers who are innovating with limited resources to solve local problems.
The Deep Learning Indaba, launched in 2017, now has chapters in 47 of Africa’s 55 nations. Shakir Mohamed, co-founder of the Indaba and a senior research scientist at Google DeepMind, sees localised AI solutions as essential to Africa’s development. “Building and owning AI solutions tailored to local contexts is crucial,” he said. “Africa has an immense pool of young talent ready to drive this vision forward.”
Hurdles on the Road to AI Development
Despite its promise, Africa’s AI ecosystem faces considerable challenges. Funding is one of the most significant barriers; building AI systems is costly, and many African universities lack the financial resources to develop training datasets, especially for languages spoken predominantly in Africa. Internet infrastructure is also limited in many regions, restricting the ability of local developers to access and deploy advanced AI technologies.
The regulatory landscape remains another obstacle. While a few African nations, including Senegal, Nigeria, and Egypt, have adopted AI strategies, a cohesive, continent-wide policy is still a distant goal. The African Union Development Agency published a white paper outlining a vision for AI, yet this was met with resistance when the African Union Commission released a competing strategy. This divergence highlights a lack of unity and a struggle for control over Africa’s AI agenda.
Harnessing AI to Address Local Needs
Africa’s AI community is already achieving notable successes. In Kenya, AI-powered applications are helping farmers identify and treat crop diseases, boosting agricultural productivity. South African researchers have developed an app using large language models to facilitate sensitive discussions around HIV, an ongoing public health challenge in the region. Meanwhile, Awarri, a Nigerian startup, is building a large language model to support Nigerian languages, with backing from the government.
Initiatives like these underscore the continent’s drive to develop AI solutions that meet local needs. Kathleen Siminyu, a Kenyan computer scientist researching natural language processing for Kiswahili, sees this momentum as unstoppable. “We’re seeing a wave of talented people with foundational skills who are now specialising in AI,” she said. “It’s a movement that’s only growing.”
Language Barriers in AI Development
One of the greatest hurdles to developing Africa-centric AI lies in language. Africa is home to a third of the world’s languages, yet many of these are oral and lack the written resources necessary for AI training. During the Indaba, Nigerian computational linguist Ife Adebara highlighted this challenge by asking how many people in the audience could write a bachelor’s thesis in their native African language—none could. This points to an issue facing AI developers: the lack of written data in African languages limits the creation of AI tools that could better serve local communities.
Despite this, there are promising efforts to address the language gap. Startups like Lelapa AI, co-founded by Jade Abbott, are developing models and datasets for African languages. This year, Lelapa AI released InkubaLM, a language model that can handle multiple African languages, offering a glimpse into a future where AI can communicate in languages like IsiXhosa, Yoruba, and Swahili.
Balancing Local and Foreign Influence
The African AI community also faces questions about the influence of foreign funding. Half of the Deep Learning Indaba’s budget comes from international sources, including Google DeepMind, Apple, and Meta. While these funds enable the conference to provide training and resources, critics argue that reliance on foreign tech companies could compromise the event’s agenda.
“There’s no question that we need more African government involvement,” said Seydina Ndiaye, a programme director who contributed to the AU’s AI strategy white paper. Others, like Timnit Gebru of the Distributed AI Research Institute, caution against allowing external pressures to push Africa into adopting AI across every facet of society without careful consideration of local needs. “AI should not be treated as a cure-all for Africa’s challenges,” Gebru said, highlighting the need for a balanced approach.
Barbara Glover, a programme manager for the African Union Development Agency, believes that African experts must drive the continent’s AI strategy. “We are capable, as Africans, of setting our own AI agenda,” she stated. As the AU seeks to reconcile differing strategic visions, the hope is that African countries can work together to develop policies that prioritise local interests over global tech giants’ agendas.
The Road Ahead
Africa’s AI journey is far from straightforward, with challenges ranging from funding and infrastructure to language barriers and regulatory disputes. Yet the spirit of innovation among African AI developers is undeniable. Through collaborative efforts, such as the Deep Learning Indaba and Masakhane’s work on language data, Africa is laying the groundwork for a homegrown AI sector that addresses local needs and taps into the continent’s vast talent pool.
The African AI landscape may be complex, but it is evolving. With continued support, Africa’s researchers and entrepreneurs are positioning the continent to become a formidable player in the global AI arena—on its own terms.
Source: https://www.technologyreview.com/2024/11/11/1106762/africa-ai-barriers/
