African startups are entering the AI race on their own terms.
From Nairobi to Lagos to Johannesburg, local entrepreneurs are building open source models to create AI applications tailored to African contexts, giving the continent a measure of data sovereignty and control.
The global AI market is dominated by a small number of giant technology companies from the US and China. Africa, with its history of leapfrogging the technological ladder, is becoming a testing ground for bottom-up, decentralized, local alternatives that could serve as a model not just for the Global South but for the rest of the world.
“There will be a global wave of private models and more open source models.” Toffen Kama said Mercy Corps Ventures, which tracks and invests in cutting-edge technology trends in emerging and frontier markets. impact alpha. “But there’s also a big wave coming from Earth, and smaller micromodels that can do the work in a more efficient way only address very narrow problems.”
Such local sovereignty means lightweight AI platforms and applications that work in areas that lack reliable energy, internet connectivity, and computing power. A language model that includes hundreds of African languages ignored by the dominant Western model. In Kenya, jacaranda health delivers pregnancy and postpartum advice to women via text message. In Senegal, Torvi Analyze weather data to help farmers optimize food production. African startups are also leveraging AI to enable access to credit, insurance, and other financial services for marginalized communities.
Policymakers and investors are discussing Africa’s huge opportunities and sudden risks at the Africa Forward Summit in Nairobi this week. The impact of AI and digital infrastructure, and how its benefits will be shared, has dominated other recent African forums as well.
“As the world and technology change rapidly, it would be a mistake to think that we can achieve growth, prosperity and development by sitting on the sidelines, and we would be missing out on hugely important opportunities.” mike mompi said Nairobi-based Enza Capital. impact alpha In line with the recent African Venture Capital Association Summit. Enza has helped more than 60 companies, including AI companies, use technology to solve large-scale, important problems across Africa. The third fund is about halfway to its $60 million to $80 million goal and is focused entirely on AI startups.
“People say the train left the station, or AI did it, or Silicon Valley figured it out,” he added. “That’s not true. They’ve built very good technology that’s relevant to the United States and the world. But that doesn’t mean Africa can’t build around technology.”
The need for local data was underscored by a Kenyan High Court ruling last year that blocked a five-year, $2.5 billion medical cooperation agreement with the United States to support programs and disease surveillance in HIV/AIDS, tuberculosis, malaria, maternal and child health. The agreement was blocked by the High Court on the grounds that it violated several health data protection rights. Ghana rejected a similar health deal, citing data sovereignty concerns. Zambia rejected the deal, citing data concerns and the need to dissociate from minerals.
Nevertheless, there is a marked increase in AI optimism in Africa. In a Pew survey conducted last fall, respondents in Nigeria and Kenya were far more excited about the promise of AI than those in more developed Western countries. In the United States, there is a growing populist backlash against AI and the large-scale expansion of data centers to power it.
Some Africans are convinced that AI will create, not destroy, jobs for the growing number of young people. Additionally, the dissemination of knowledge and expertise enabled by AI will increase productivity in farmers, healthcare, and other sectors, driving economic growth.
“I think there will be a lot of efficiency gains,” Mompi says. “We’re looking at GDP per capita, increasing with innovation and efficiency,” which in turn will drive growth, he says.
AI FOREVER
Many of the AI-based applications being developed in Africa are designed for low-connectivity computing environments. Startups that solve key problems for marginalized communities will be able to iterate and bring their products to market faster. AI coding solutions have reduced production costs, including high salaries for software teams.
Mercy Corps Ventures runs an accelerator program focused on AI for climate and financial resilience. With both programs, Mercy Corps seeks to support startups that combine AI and satellite imagery to predict droughts and other climate shocks, information critical to providing parametric insurance.
Its portfolio includes U.S.-based flood base, Provides AI-powered satellite analytics for flood mapping and supports parametric insurance. In partnership with the insurance sector of the African Union Africa’s risk capacity We plan to develop parametric insurance products in 2023. Senegal’s Tolbi uses AI to analyze and provide data on weather patterns, irrigation requirements, soil conditions, and crop needs across French-speaking Africa.
In Kenya, Jacaranda Health, a nonprofit organization that partners with the government to provide access to quality maternal care and reduce infant mortality in low-resource areas, deployed AI capabilities with government support. rockefeller foundation. The service, called “Promoting Pregnant and Postpartum Mothers through SMS (PROMPTS),” acts as a two-way SMS exchange in Swahili, answering mothers’ questions through the different stages of pregnancy and also providing postnatal advice.
based in california digital green Launched FarmerChat in partnership with Open AI and support from the Rockefeller Foundation to give farmers in Kenya, India, Nigeria, and Brazil access to agricultural advice and government resources in local languages through a messaging app whatsapp and telegram.
Based in Kenya Crinivais complementing its two women’s mini-clinics in Nairobi with an AI chatbot embedded in Whatsapp. The company built its services on OpenAI’s open resources and Google’s open resources. Medgemaa collection of open models for health AI development. Cliniva received the grant as part of the Chat for Health and AI Accelerator cohort that launched last year. johnson & johnson foundation and mulago foundation, Turn.io OpenAI helps innovators expand access to healthcare in low-resource settings using AI-enabled chat services.
Kriniva’s Yulia Sidorova She says she is concerned that AI models trained on patient data from around the world do not capture the reality and needs of patients in Africa.
“What worries me is that my patients don’t look like foreign models at all,” she says. “This is not just a question of fairness: a pregnancy in rural Kenya is clinically very different from a pregnancy in London. [AI models] Based on local data, it could be dangerous. ”
At the moment, “developing local models is very expensive, and in the end it’s always a question of economics,” Siderova says. Cliniva’s strategy is to continue building on the existing open model by introducing its own guardrails, memory, and proprietary data.
five layer cake
If Africa’s first wave of AI startups focused on applying the technology to persistent challenges, from food production to financial services to healthcare, the second wave is focused on AI infrastructure itself, from data centers to large-scale language models and the data used to train them. (look, “Investing across the technology stack to orchestrate “better AI””). Nvidia’s jensen fan describes a five-layer cake that includes energy, chips, physical infrastructure, and AI models and applications.
Building in an inclusive way is key. According to the Global Center for AI Governance, which published a report on the AI landscape in Africa last year, open source AI can help democratize AI innovation and reduce reliance on expensive proprietary technology.
Community-specific AI projects Funiki Federationis a collaboration of African language technology startups that helps communities manage their data, goals, and deployments. “We don’t want to build a one-size-fits-all model,” he said. Timnit Gebru Professor at the Institute for Distributed Artificial Intelligence at South by Southwest in Austin earlier this year.
“I want to create many models of different people from all over the world, because there is more than one way to be human.”
The Rockefeller Foundation is promoting the creation of an African database to accelerate the localization of AI. This allows companies like Jacaranda Health and Digital Green to offer services tailored to marginalized populations in local languages, rather than generic services.
“What’s top of mind is the need to actually build datasets, not only because that data is important so that it can support AI and large-scale language models, but also because it helps the community and others understand the continent,” Rockefeller said. Elizabeth Yee Say. “At the end of the day, there is no finance without data, so the two need to work together.”
In addition to our broader digitalization efforts, incorporating some of Africa’s 2,000+ languages into our AI models is a key focus. Based in Johannesburg Relapa AIFor example, we are building models for African languages such as Swahili, Yoruba, IsiXhosa, Hausa, and IsiZulu, which have more than 350 million speakers. Its “small language model” is designed for real-world situations with limited computing power, cloud infrastructure, and connectivity.
Masahaneis a grassroots organization whose name means “building together” in isiZulu and works to empower Africans with AI models that incorporate cultural context and nuance.
Building and scaling AI in Africa will also help the continent protect and manage its highly-demanded resources and data. Without input and innovation from Africa, Enza’s Mompi warns that AI could become a new driver of exploitation. “The world is adding undervalued products to the outside world and selling them back.” [to Africa]” he says.
Hakiman Enza portfolio company, is building a language model trained on unpublished African court cases to enable speeding up legal research procedures slowed by unstructured information. Volo Earth Ventures Based in the US and South Africa rebellious AIcompressing large language models to reduce AI energy usage.
“Our market is fragmented. There’s a lot of data that doesn’t exist systematically,” he continues. “There is a lot of work to do to structure data, incorporate local languages, build platforms on African foundations, and solve problems locally.”
