How Facial Recognition Technology Is Different In Africa
There's a reason many Africans are wary of the identification technology: It doesn't work as well for people with dark skin. That's where Charlette N'Guessan, a young Ivorian researcher, comes in.
ACCRA — She's not ashamed to say it: The coronavirus pandemic has been "a very good thing" for Charlette N'Guessan. The same goes for Africa's forward-looking tech entrepreneurs in general.
"With the challenges posed by COVID-19, the continent is waking up," the young Ivorian says. "People are thinking innovation, ideas, change. This crisis gives credibility to what we are doing."
It must be said that N'Guessan's chosen field — facial recognition — is largely unexplored in Africa. It also arouses a fair amount of suspicion, and for good reason: Existing algorithms, including the best out there, are less accurate at identifying individuals of color, as tests conducted in the United States have revealed an error rate five to 10 times higher for these populations.
It was partly to correct these biases that the 27-year-old N'Guessan joined forces in 2018 with three other computer engineers. They met at the Meltwater Entrepreneurial School of Technology (MEST) incubator in Accra, Ghana, where she was undergoing training in coding and entrepreneurship.
The start-up they founded has developed its own software, Bace API. To ensure that it would perform well with dark skin tones and be adaptable to the local market, the team relied on a very diverse data set, including a large sample of sub-Saharan African faces.
"In the beginning, we even practiced on the other members of the incubator," N'Guessan recalls with a laugh.
Cybersecurity is a problem everywhere in Africa and even more so in the financial sector.
The development of this solution is intended to respond to very concrete issues. In 2017, cybercrimes cost African economies $3.5 billion, according to the Kenyan-based consulting firm Serianu.
Ghanaian financial institutions are facing a massive problem of identity theft, and are losing hundreds of millions of dollars per year as a result, according to N'Guessan and her colleagues. To help combat the problem, Bace API provides banks and FinTech companies with a system to verify the identity of customers remotely using "live" (moving) photos to ensure that the person is real and not a robot.
This achievement earned N'Guessan the Africa Award from the Royal Academy of Engineering, a prestigious British institution that issues annual prizes for innovation. The recognition provided a welcome boost for the start-up, as did the grant money (nearly 28,000 euros) that came with the award.
Born in Abidjan, the Ivory Coast's largest city, Charlette — "with an e, not an o," she says — grew up in the working-class neighborhood of Yopougon. Her father was a math teacher, and she has five sisters, all with the same first name... Charlette! "But each of us was given a middle name," she explains.
N'Guessan at the Meltwater Entrepreneurial School of Technology (MEST) incubator— Photo: Charlette N'Guessan Twitter
"I've always been encouraged to follow my path and to dream of great things," she says. "Probably because we were only girls at home, my father didn't see why we would have less interesting career plans than boys."
After studies in electronics and computer networks and internships in companies in the Plateau, Abidjan's business district, Charlette N'Guessan was selected to train at the renowned MEST incubator. She is one of the few French speakers to join.
It is in Ghana that she chose to focus on facial recognition, a discipline derived from Artificial Intelligence (AI). She says this decision was a gamble, as the technology, perceived as discriminatory, is the subject of much debate in Africa. What's more, fingerprint identification is already well ahead in many African countries. But the health crisis is changing that.
"Today, with what we have just gone through, everyone wants to have access to their services remotely," says N'Guessan.
She sees prospects in many areas: from education (with the implementation of online testing platforms) to individual passenger transportation (to facilitate the hiring of drivers) and public services, such as electoral processes that require extensive voter registration.
"We need more "made in Africa" solutions instead of products from elsewhere"
At present, Africa is lagging behind in the creation of start-ups and AI technologies. A Stanford University report suggests that as of 2018, the bulk of AI investment was concentrated in 20 countries around the world. Not one of those countries is in Africa.
Nevertheless, a few centers of excellence are beginning to emerge there. In Ghana, Google opened its first African AI research laboratory in early 2019. And in Kigali, the capital of Rwanda, the African Institute for Mathematical Sciences has launched a master's program dedicated to machine learning and AI, in partnership with Facebook and Google.
"We need more "made in Africa" solutions instead of products from elsewhere," says N'Guessan.
This plea echoes fears expressed across the continent, where people are concerned in particular over China deploying its surveillance technologies. In Zimbabwe, for example, the government has signed a cooperative agreement with CloudWalk Technology, the Chinese leader in the sector, to implement facial recognition on a large scale. And Huawei has developed numerous partnerships with various African countries to develop its "safe city" initiative, a project to "secure" cities through installing smart cameras.
For "ethical" reasons, Charlette N'Guessan does not want to venture into this field of public and police video surveillance. But there are plenty of other areas, she believes, where technologies developed by and for Africans can be of great use.