I spoke at Fortune’s Brainstorm Tech conference on AI at the frontier. Below is a fleshed-out version of my talk. You can watch the video of the talk at Fortune.com.
Today I want to talk to you about a frontier market and the opportunity it holds for AI. Before I start, I need to make a clarification. When I speak of AI what I am talking about is, in fact, a blend of computer science and statistics that take into account human utilities. These are systems that perform planetary-scale-inference-and-decision-making, for example, google maps.
You probably have come across the Africa rising narrative that has been going around for a few years now. According to the IMF, sub-Saharan Africa will be home to several of the world’s fastest-growing economies, 60% of its 1.25 billion population are under the age of 25. By the end of the decade Africa will have 90 countries with at least a million residents, Europe, by contrast, has 17, while the US has 10. Further, half of all global mobile payment accounts originate in Africa, that is 120 million active accounts.
When I speak of machine learning at the frontier, I am not speaking about the frontiers of knowledge; instead, I am speaking of the applications of these planetary-scale-inference-and-decision-making systems in a frontier environment like Subsaharan Africa. If you look up the dictionary definition of a frontier, you will see that it is defined as the extreme limit of settled land beyond which lies wilderness. One of the most popular frontiers in our present imagination is the American West before the Pacific settlement. That period laid the foundation of what would later become Silicon Valley. Today we are the edge of another frontier, a present technological wilderness which is presently being transformed through technology.
There is something powerful about using the same habit primitives. Uber is Uber whether I am in Lagos, Nairobi, or Paris. Google Maps is Google Maps, whether I am in Aspen, Capetown, or Accra. GitHub is GitHub, whether in am in Andela in Lagos or GitHub HQ in San Francisco. All these primitives are the same, no matter where you are in the world. What is unique about the African perspective is the opportunity to create something altogether new. What do you do when the roads are bad, or there are no roads? You take to the air. There is a logistics company out of San Francisco called ZipLine; what they do is deliver blood products via drones to rural areas in Africa. The drone solution to logistics is an example of the kinds of innovations that uniquely flourishes in the African environment.
The great thing about the Africa Frontier in terms of the future of technology and particular machine learning is that it is a very rich source of data. Africa has over a billion people, eager and excited to join the rest of the world online. Many are already online from Nairobi, to Lagos, to Addis Ababa and Accra, to name a few. Last summer, I went to Kenya for a holiday. I arrived in Nairobi, where immediately I was able to get a cellular data plan via Safaricom. I was astonished that the app that came with my mobile data plan already determined some baseline of my creditworthiness and was able to extend some financial services to me. From Nairobi, I went with my family to Masai Mara on the border with Tanzania. Through every single small village we drove through, we could always have access to mobile money. This deep penetration of mobile payments technology from the urban to the rural was an eye-opening experience for me as a person who does ML for a living. We, the practitioners, often don’t get to truly understand the power of the technologies that we build until we are forced to rely heavily on our products to shape our experiences. One such learning opportunity presents itself when we travel. Transfering decision making to our data products presents itself during travel to a new place. It is then that we get to see and appreciate the power of machine learning.
In my opinion, there are a few things necessary for lifting people out of poverty; these include respect for the rule of law, access to capital, and access to education. In the last century, many African countries changed their governance structures; many went from colonialism to democratic systems. These changes weren’t always smooth; many of the emerging democracies had a difficult time meeting the demands of governance; as a result, many Africans became used to being somewhat self-sovereign individuals. Many of us came of age in countries where our families had to provide our power, education, water, and often, healthcare.
As difficult as these conditions were and in some cases are, they created a mindset in the population that makes it open to permissionless and decentralized ways of thinking. I believe this is why we see the adoption of cryptocurrencies and blockchain technologies in the region. For example, one of Binance Labs five hubs is in Lagos, Nigeria. Payments are one of the last major technological infrastructure keeping Africa in the dark. Decentralized systems of finance open the opportunity for the vast majority of the unbanked to participate in the financial economy we all take for granted. In a place with no credit score, it is the machine learning folks that build the systems to learn new ways of deducing creditworthiness. Blockchain technology with the transparency that it promises means that in a place rife with corruption, we can drastically reduce its impact by bringing many processes online, from standardized testing, to payroll, to procurement and so forth.
I want to end by telling you a data story in the African context. Last month, I went on a factfinding mission with our GitHub CEO, Nat Friedman, and his Chief of Staff, Naytri Shroff. Looking at our state of the Octoverse report, our interest in Nigeria heightened based on its outsized position as one of the top places where GitHub is growing. One of the activities we did while in Lagos was hosting a GitHub open-source mixer. We took a data-driven approach to curate the list of open-source contributors we invited. We ran a query on our dataset to determined the top 50 open-source developers on GitHub from Nigeria within the last 60 days.
When the developers arrived, many wondered why they had been invited. We told them it was because of their contributions to open-source.
Every developer earned their way into that room; their entry wasn’t based on who they knew, what family they were born into, or whether they were rich or poor. Even though I theoretically knew that data-driven approaches could be a way to mitigate the impact of bias, I had not truly appreciated the profound effect it can have. We brought this community together for one night to say thank you and to learn.
That GitHub Open-source developers mixer showed every developer in that room that their voice matters, that their work matters, that their gift of open-source to the world matters. The open-source mixer gave me a glimpse of a different kind of future; a future we can build together with the power of all of us with the power of each of us.
Notes
- For a thorough understanding on planetary-scale-inference-and-decision-making systems, consider reading Michael I. Jordan’s essay on AI, Artificial Intelligence — The Revolution Hasn’t Happened Yet.
- For a brief exposition on Africa’s growth potential, consider reading this report from the Brooking’s Institute.
