The Growth of Artificial Intelligence in Africa: On Diversity and Representation

Published on May 19, 2021

Share this post

by Mohini Baijnath, Neil Butcher and Associates (NBA)

As the world ushers in the Fourth Industrial Revolution (4IR) – which is characterized by increasingly blurred lines between the digital, biological, and physical worlds[1] – technologists are coming to grips with the opportunities of emerging technologies such as Artificial Intelligence (AI), robotics, and the Internet of Things. These and other emerging technologies offer exciting possibilities; in theory, they might allow us to galvanize unprecedented socio-economic change and democratise access to services such as the internet, education, and healthcare.

AI is poised to impact Africa in several ways. It has been hailed by many as a transformative force for African societies, promising to reduce inequality, alleviate poverty, and improve access to public services like health and education. But deployment of these powerful technologies is still in a fledgling phase on the continent, and there are significant challenges to overcome in building capacity to fully harness their potential. Critical among these is a lack of diversity in the field globally, which filters into all areas of AI, from developing datasets to systems development and deployment. In this context, diversity refers to the practice of including people from varied socio-economic, racial, ethnic, and cultural backgrounds, as well as people of different genders and sexual orientations.

The newly launched International Research Centre for AI (IRCAI) with the involvement of IRCAI personnel has just released a report on AI in sub-Saharan Africa, which explores a capacity building agenda for AI in the region. NBA had the privilege of collaborating with Canada’s International Development Research Centre (IDRC) and the Knowledge for All Foundation (K4A) in developing the report under the IRCAI banner.

The report examines the AI landscape in sub-Saharan Africa, focusing on three key stakeholders involved in AI capacity building in the region, namely, Centres of Higher Education and Training, Governments, and the broader AI community. It draws on key findings from multiple data sources. New primary research was conducted with stakeholders, which provided data about AI in the region that was not previously available.

In the study, 58% of respondents representing the AI community noted that diversity was an issue in AI in their country, institution, or organization – particularly gender diversity. The study also found that there are more males than females involved in AI-related activities in Centres of Higher Education and Training. For example, within research and development activities, the gender breakdown for students was 77% male and 23% female and for staff it was 71% male and 29% female.

This data, along with an extensive desktop review, was analysed to generate recommendations for each of these stakeholders. One focus areas in this regard is diversity. The report provided the following recommendation:

Use AI and related technologies as an opportunity to create and reinforce diversity. Key to this will be to facilitate and promote skills development of diverse people and to make concerted efforts at levelling the playing field for women and other minorities in the industry.

There have been several reports documenting the lack of diversity in AI around the world, a problem that perpetuates biases of various kinds and disregards the potential contribution that different people can make to the field. Examples include a new tool for creating realistic, high-resolution images of people from pixelated photos which exhibited racial bias by turning pixelated yet recognizable photos of well-known people of colour into high-resolution photos of Caucasian people. A second example comes from the healthcare sector where, for decades, there has been a disproportionate amount of data collected from male subjects, which might mislead AI into determining that men predominantly suffer from diseases such as cardiovascular disease. Central to developing and maintaining diversity within the field of AI is ensuring that the humans who build and deploy these technologies come from varied backgrounds and offer diverse world views. This can and should be intentionally infused in the field of AI by, for example, assembling more diverse design teams, reviewing recruitment procedures for those involved in AI-related jobs, and supporting NGOs whose activities drive the diversity agenda. Failing to actively support a diverse workforce will perpetuate inherently deficient AI that sustains gender, racial, class, and cultural biases.

This diversity should also include representatives from around the world instead of just more technologically advanced countries that currently dominate the sector. The report found that there are ‘promising signs of AI ecosystem growth in [sub-Saharan Africa]’ (pp44). This is corroborated by recent mapping of emerging AI hotspots in African countries, which identified a total of 149 players in sub-Saharan Africa (SSA). However, it also pointed out that there are alarming signs of global South exclusion. For example, at the 2016 NeurIPS conference – a global conference dedicated to artificial neural information processing systems – no papers from African countries were accepted. Likewise, over 100 researchers were refused travel visas to travel to Canada for the 2018 NeurIPS conference. Although there is strong evidence of AI Communities of Practice forming in Africa, it is crucial that the continent joins the global conversation.

Globally, society and business can benefit from more purposeful focus on diversification because it significantly improves the quality of datasets. For AI to deliver accurate outcomes, datasets need to be representative of populations. Accurate data not only provides reliable insights, it also presents opportunities for businesses to capture new markets and for governments to deliver public services where they are most needed. Moreover, representation is more than a box-ticking exercise; it is a moral and ethical imperative, acknowledging the valuable contribution that a diverse set of individuals can provide.

The report concluded that the process of building a vibrant AI ecosystem in SSA offers an opportunity to ‘employ AI and related technologies to facilitate and reinforce equality through skills development and concerted efforts to level the playing field for marginalised groups in the sector’ (pp85). Because AI is still gathering momentum in sub-Saharan Africa, it is possible to design diversity mechanisms at the outset. Ensuring diversity at a systemic level will have a decisive, lasting effect on the sector.

AI is powerful enough either to reduce or reinforce societal stratification, according to our choices. Key to ensuring that the latter happens is to hold Centres of Higher Education and Training, Governments and the broader AI community accountable for infusing the sector with a diverse and representative range of people. More than this, it is about promoting representation by highlighting the opportunities that diversity presents through new voices that provide innovative solutions to greater numbers of people.

You can access the full report here.

[1] Ndung’u, N. and Signé, L. (2020). “The Fourth Industrial Revolution and digitization will transform Africa into a global powerhouse.” Brookings Institute.


Presenting the Global Top 100 outstanding projects: Artificial Intelligence-Based Referral System for Patients With Diabetic Retinopathy in Jalisco

Presenting the Global Top 100 outstanding projects: Artificial Intelligence-Based Referral System for Patients With Diabetic Retinopathy in Jalisco

This project provides an inspiring example of an AI-based referral system for patients with diabetic retinopathy. Aware of the high prevalence of diabetes and a shortage of ophthalmologists and eye illness detections services in the Mexican State of Jalisco, the founding members of this
project developed an AI model based on deep convolutional neural networks (DCNN) to enable a timelier detection of Diabetic Retinopathy in people’s eyes. The trained models preprocess and evaluate the suitability and quality and classify relevant ocular documentation with an exceptional specificity and sensitivity on local data.

Presenting the Global Top 100 outstanding projects: MALENA

Presenting the Global Top 100 outstanding projects: MALENA

MALENA is an NLP-based ESG analyst tool enabling investors to quickly review ESG texts developed by the International Finance Corporation (IFC) to make SDG-integrated investments in emerging markets. The tool is able to identify more than 1,200 ESG risk terms, then performs a sentiment analysis to assign positive, negative or neutral sentiment to the risk terms, depending on the context.


International Research Centre
on Artificial Intelligence (IRCAI)
under the auspices of UNESCO 

Jožef Stefan Institute
Jamova cesta 39
SI-1000 Ljubljana


The designations employed and the presentation of material throughout this website do not imply the expression of any opinion whatsoever on the part of UNESCO concerning the legal status of any country, territory, city or area of its authorities, or concerning the delimitation of its frontiers or boundaries.

Design by Ana Fabjan