Crowdsourcing Artificial Intelligence in Africa: Findings from a Machine Learning Contest

In this paper, we study the crowdsourcing of innovation in Africa through a data science contest on an intermediated digital platform. We ran a Machine Learning (ML) contest on the continent's largest data science contest platform, Zindi. Contestants were surveyed on their motivations to take part and their perceptions about AI in Africa. In total, 614 contestants submitted 15,832 entries, and 559 responded to the accompanying survey. From the findings, we answered several questions: who take part in these contests and why? Who is most likely to win? What are contestants' entrepreneurial aspirations in deploying AI? What are the obstacles they perceive to the greater diffusion of AI in Africa? We conclude that crowdsourcing of AI via data contest platforms offers a potential mechanism to alleviate some of the constraints in the adoption and diffusion of AI in Africa. Recommendations for further research are made.