In conjunction with our AI4H call, we took the opportunity to learn more about AI from our AI Sector Lead, Deogratias Mzurikwao.
Artificial intelligence (AI) is a program that mimics the working principle of human intelligence. Its practical application is being felt in many different industries, but the potential of artificial intelligence in healthcare is vast. It can be used to interpret medical imaging, such as CTs and MRIs; develop personalised medication, looking at an individual’s gene expression to determine which medications will be effective for their particular body; or analyse large amounts of data to determine epidemiological trends (among many other applications). In Africa, this is still an emerging field, and although there are examples of AI interventions functioning on the continent, very few (if any) are locally developed.
AI has great potential to increase the accessibility, accuracy, and speed of diagnosis and the effectiveness of treatment. It is especially suited to the African context, where the shortage of well-trained healthcare workers continues to be a serious challenge.
In Tanzania, for example, there are fewer than 20 oncologists, most of whom are located in Dar es Salaam and many of whom are in management positions and don’t see patients. Because of the many barriers to seeing an oncologist, more than 80% of breast cancer cases in the country are diagnosed in later stages. More than 50% of women diagnosed with breast cancer in Tanzania will die as a result of late diagnosis. The potential for AI means that a woman would be able to be screened regularly, without having to travel to see a doctor and before they are experiencing any symptoms.
As with most emerging technologies, the high potential reward of AI also comes with risks. Because it is at such an early stage, there is currently very little regulation and very few policies on AI in Africa. The primary risk of unregulated or poorly done AI is bias. Some algorithms will favour one output over another, which can lead to misdiagnosis or other mistakes and, in healthcare, a small error can be a matter of life and death. Additionally, sometimes the algorithms behind AI are black boxes and it’s not possible to determine how they make decisions, meaning their efficacy and accuracy cannot be determined.
Many of the AI innovators in the private sector are acquiring their knowledge from online resources, not universities. Very few universities in Africa offer AI-related courses, but there are brilliant young people who access online resources and teach themselves. This has its risks, however, because the online resources often don’t train on ethics, the importance of knowing how data is obtained, ensuring they are using quality data, whether the privacy of data is protected, and so on. The technical use and application of AI is one thing, but the ethical use of it is just as important. Many of the online resources show how amazing and useful AI is, but they don’t show the dark side of AI and how it can fail.
“The young people who are teaching themselves from online resources, they have this perception that AI is like a hammer and whenever I see a nail I just hit it.” - Dr. Deogratis Mzurikwao
Responsible AI considers where you're collecting your data set from. Did you consider gender balance? Age balance? Is your model explainable? Can you explain to a medical doctor how the decision was made? We shouldn’t wait until we see AI causing harm on the continent before implementing regulations; we have to regulate it early. In addition to the harm that could be caused, the worry is that if one or two AI products fail the market, governments might rush to implement harsh restrictions or even simply ban AI altogether. With early regulation, anyone coming into the market can understand the guidelines and implement their solution accordingly, following the correct steps.
Villgro Africa is committed to promoting AI innovation in healthcare that is responsible, safe and effective. Following the AI for Development (AI4D) call in 2021, we are excited be currently launching a second cohort, AI for Health (AI4H). (Applications are due on the 11th of November, 2022.)
The initial AI4D call gave us a deeper understanding of what is happening in the ecosystem, with applications pouring in from across the continent. We learned that many AI grants are going into academic research rather than into the private sector even though, in many African countries, innovations often come from the private sector rather than academic institutes, which focus mostly on research. We also saw that AI innovation is lacking significantly within certain demographics, including women and Francophone Africa.
The AI4H call, in partnership with IDRC, AI4D, the Africa Oxford Initiative, I-DAIR, Johnson and Johnson Impact Ventures and NVIDIA is looking for health-focused African-led AI innovations. We hope to have more applicants who are female and come from Francophone countries, and we will provide training based on the challenges we have identified in our previous call. We also hope to have more data sets because part of the plan is to convince these companies to use local data sets and then to keep that data publicly available. This means that any player on the continent would be able to use locally, publicly available, ethical datasets.
AI presents unique challenges. Along with the risks and the fact that it is new on the market, AI is different from other innovations. Even if someone has AI skills, they still have to collect data sets, which is a very expensive process. AI innovations require access to data, training the model, ensuring the model is working, and then going through regulatory approval. This is why we will offer a follow-on grant as a part of the AI4H program. We will give the initial grant to several chosen candidates, then provide a few promising companies with follow-on funding so that they can scale.
Dr. Deogratis Mzurikwao is Villgro Africa’s AI Sector Lead, building the capacity of and providing technical guidance to the Villgro team, our portfolio companies and the innovation sector at large. He works to create a community of stakeholders who are working in emerging technologies, and establishing Villgro Africa’s role as a leader in the implementation of AI in healthcare innovation.