Zindi, the data science competition platform established in SA, aims to make AI accessible to everyone and highlight the region’s talent pool of data scientists.
With more than 80,000 members from 190 countries, Zindi is one of the most significant communities of data and AI practitioners in the Middle East and Africa (MEA) region.
It hosts an entire data science ecosystem of scientists, engineers, academics, companies, NGOs, governments and institutions focused on solving Africa’s most pressing problems.
Business Day spoke with platform founder and CEO Celina Lee, who says investing in skills is the only way to fully harness AI’s potential and ensure that SA has the capacity to develop solutions that successfully and distinctively tackle the country’s problems.
What prompted you to start Zindi and what challenges does it aim to solve?
Two problems prompted me to start Zindi. One was a frustration that organisations in Africa, whether government or companies, seemed to assume they had to look outside Africa to get the talent and solutions they needed. And, two, in my previous work I had contact with the growing community of data scientists in Africa, and I knew that they were extremely talented and driven, but they didn’t have access to the training, the exposure, the networks, they needed to grow their careers. I like to say Zindi has African roots but global reach.
What is your overview of the AI landscape and ecosystem in SA?
It is growing rapidly. Companies are adopting more AI tools for internal operational efficiency gains as well as to improve services and products to customers. I would say larger companies are driving the former and start-ups are the ones driving the latter.
Could you share some examples of the challenges that participants on Zindi have solved?
We have done a number of challenges with organisations in SA including Absa, RMB, SAEON [the SA Environmental Observation Network], Sansa [the SA National Space Agency] and Sanral. The largest prize challenge we did for an SA organisation was with RMB, which was working on a now casting model for CPI. They wanted to take a more innovative approach to achieve greater accuracy. So we ran a challenge for them that called on participants to use alternative data such as traffic patterns, pollution, social media data and others to train a machine learning model. We had a rolling leader board where participants would use their models to predict the next month. We also awarded the winners based on who came closest to the correct CPI indicators for that month.
Where are the biggest opportunities to use AI and bolster the SA economy?
The issue is that SA is still undergoing a digital transformation. It is certainly more advanced than other countries, but still has a ways to go compared to more advanced economies. AI is fuelled by and runs on connectivity and digital data. So low-hanging fruit would have to be in areas where the level of digitalisation is high — say fintechs and the financial sector more broadly. AI can help automate many of the processes. It can make customer interactions and support faster, more immediate, and be more effective in driving desired customer behaviour. On the other hand, the biggest opportunity to use AI to bolster the SA economy is a larger question. AI can be used to extract and share market insights to help with more efficient investments in the country.
Where are the biggest risks AI presents to SA businesses, governments, and citizens?
The biggest risk is the irresponsible use of AI. The solutions being shipped today are trained on data that is mostly not South African. So when adopting an AI solution, in all contexts, but especially for businesses and people in Africa, there has to be a human review of the solution’s performance and an assessment of any inaccuracies and potential biases embedded in the solution.
The most obvious example is that a company can sell facial recognition software as “accurate”' But if it systematically performs poorly on certain segments of the population, it is, for now, the responsibility of the business that buys this software and uses it to recognise the bias and evaluate the risk of using it depending on the use case.
The risk of faulty facial recognition ranges from annoyance if it does not work to unlock your phone to life-threatening if the police adopt and rely too heavily on this software and misidentify criminals.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.