South African uses AI, digital mapping to fight land segregation
Many years down the line since South Africans gained independence and fought racial segregation, the relics of these colonial administrators are traceable in the lives of the average citizen.
One of the legacies of the apartheid system in South Africa is the physical division of communities based on economic and racial lines. This division has resulted in townships being situated near workplaces, facilitating commuting but simultaneously making it challenging for residents to access essential services.
Thus, in the post-apartheid era, the impact of this division continues to affect the socio-economic well-being of the members of low-income communities in the country.
In an attempt to correct the seeming anomaly in the country, 28-year-old Raesetje Sefala has collaborated with computer scientists Nyalleng Moorosi and Timnit Gebru at the nonprofit Distributed AI Research Institute (DAIR), established by Gebru in 2021, in employing computer vision tools and satellite images to examine the effects of racial segregation in housing.
Their aim is simple, to analyze these impacts comprehensively, with the ultimate hope that their research will contribute to efforts aimed at reversing racial segregation and fostering more inclusive housing practices.
“We still see previously marginalized communities’ lives not improving; it’s just very unequal and very frustrating, we want the work to push the government to start labelling these townships so that we can begin to tackle real issues of resource allocation,” Sefala is quoted by MIT Technology Review.
By gathering millions of satellite images of all nine South African provinces, over the past three years, the group has constructed a dataset that maps out townships, enabling them to study the evolving demographics and sizes of neighbourhoods to track improvement in the lives of residents.
They utilized geospatial data from the government, detailing the locations of various neighbourhoods and structures across the country. Leveraging this extensive dataset, the team employed machine-learning models to train an AI system capable of categorizing specific areas as affluent, non-affluent, non-residential, or vacant land, MIT Technology Review reports.
