Slowly but steadily, countries across Africa have begun preparing for the Fourth Industrial Revolution (4IR), where advances in Artificial Intelligence, Automation, Internet of Things (IoT), Cloud Computing, Robotics, 3D Printing, Nanotechnology and Advanced Wireless Technologies will radically alter the way we live, work and govern our societies. Artificial Intelligence has in particular made significant inroads in Africa, with AI enabled start-ups and other AI-focused institutions beginning to make an impact on the economy, social life and governance.

Governments in countries like Ghana, Nigeria, Kenya, Tunisia and South Africa which have made some of the most significant progress in Artificial Intelligence in Africa have been supportive of these initiatives through monetary support for AI research and development and the promotion of STEM education. Nevertheless only a few like Kenya and Tunisia have AI national strategies which can inform AI’s integration within government and public services. However for AI – a cornerstone technology of the 4IR, to make optimum impact in Africa, sweeping structural changes have to take place in the various country contexts on the continent. I explore 3 major areas below.   

Data infrastructure

Artificial Intelligence applications which solve practical problems acquire their ‘’Intelligence’’ by learning from very large datasets. For example AI models built for facial recognition will have been fed with very large datasets consisting of thousands of human faces in order to be trained on what constitutes a human face. By this token, societies and organizations with highly developed data capture, storage and processing ecosystems are better placed to optimally benefit from advances in AI.

This puts Africa at a strategic disadvantage because Africa, like much of the Global South, is data poor. In Africa and much of the Global South, public data collection for national accounts, household and firm surveys, data collection through administrative systems such as birth records, pensions, tax records, health and census are performed infrequently, and often lack the granularity necessary to make meaningful inferences about small, sub-populations of interest. And where some data exist, they are often not in digitized, machine readable formats which can be immediately harnessed for AI applications.

Therefore, in the public sector where AI applications might have been applied for the greatest public good on the continent, the data infrastructure is sadly non-existent or severely inadequate. In a show of what’s possible in a well developed data ecosystem, the National Health Service of the United Kingdom (UK) has collaborated with Google to bring the benefits of AI to public health in the UK through schemes such as rapid detection of cancers. This rapid diagnosis is powered by training AI models with large datasets of patient data within the NHS system. 

It is not a surprise therefore that some of the most promising AI applications in Africa are almost entirely private sector driven. Private sector organizations in Africa typically have data that is collected cost-effectively, with high frequency, and at fine levels of granularity. These include data from mobile phones, electronic transactions, social media, health and fitness apps and satellites which have driven the continents advances in AI applications such as chatbots and AI virtual assistants. However for Africa to fully tap into the potential of its emerging AI economy, the next decade must be focused on developing her public data ecosystems, and possibly effectively integrating them with the private sector in ways that stimulate development and protects human rights.  

Employment and Economic shifts

Experts are not unanimous on the effects of AI and automation on the future of work globally. There is a school of thought which states that the aggregate productivity gains across all economic sectors brought about by advances in AI will even out any initial job losses occasioned by AI and automation. Other prominent thought leaders describe a more sombre outlook for the future of jobs and labour. However one thing they all agree on is the outsized effect advances in AI and automation will have on the future of jobs in Africa, compared to other parts of the world.

Sub-Saharan Africa is already the world’s youngest region today with more than 60% of its population under the age of 25. By 2030, the continent will be home to more than one-quarter of the world’s total under-25 population, growing the size of its workforce by more than the rest of the world. Nevertheless, World Economic Forum data reveals that African countries are very vulnerable to job displacements occasioned by AI and automation. The statistics below illustrates the vulnerability:

• Sub-Saharan Africa exhibits a high-skilled employment share of just 6%, a contrast to the global average of 24% as South Africa, Mauritius and Botswana lead the way in the local availability of high-skilled jobs while others, such as Ethiopia and Nigeria, maintain large proportions of workers in lower-skilled jobs – which are more susceptible to automation. 
• From a technological standpoint, 41% of all work activities in South Africa are susceptible to automation, as are 44% in Ethiopia, 46% in Nigeria, 48% in Mauritius, 52% in Kenya and 53% in Angola. 

In light of Africa’s vulnerability to extensive job displacement possibly occasioned by AI and automation, urgent steps need to be taken to implement a bottom-up revision of curricula in schools across Africa. More than ever before, industry participation and input is needed in re-shaping learning and instruction in educational institutions to make ready a workforce for the rapidly changing workplaces of the 21st century. What has been observed so far seems more like a top-bottom approach largely led by the private sector, with the establishment of AI research centres across Africa by the global technology giants. Google opened its AI lab in Accra in April 2019, and the Africa Institute for Mathematical Sciences was established in Kigali Rwanda in 2016 to provide high level manpower in AI and machine learning for Africa.

A more deliberate bottom-up approach will require governments to fashion out policies which respond to the changing nature of employment on the continent, setting the agenda for decades ahead. Implementing this policy might involve tactical steps like investing more in STEM education right from primary, secondary or tertiary education levels. Nevertheless any action needs to flow from deliberate policy which guides government efforts, rather than uncoordinated, knee-jerk government responses to the problem.   

Human Rights and Accountability

All over the world, developments in AI have outpaced human rights considerations in the design and implementation of these systems. Only belatedly have corporations at the forefront of AI development given serious thought to human rights and accountability in the implementation of AI systems, often in response to pressure from civil society. Both by design and function, AI systems have the potential to hurt human rights, and I explore two areas in the African context where AI systems can do the greatest hurt to human rights.

Data privacy abuses are among the most important ways AI systems can be used to hurt human rights. AI systems need to be trained on massive amounts of data in order to function effectively, and in Africa where only about 23 countries have data protection laws, and even fewer (9) have data protection authorities, it is easy to see the potential for data privacy abuses for AI applications which interface with personal identifiable data of citizens, not least financial and health information.    

Another outlet for human rights violations comes in the shape of the roll-out of facial recognition technology across major cities on the continent. In response to a report by the Wall Street Journal which asserted that Huawei technicians had helped intelligence officials in Uganda to spy on their political opponents, Ugandan police confirmed that the technology company Huawei is rolling out a massive surveillance system that uses facial recognition and other artificial intelligence software to fight crime in the country. Opposition figures within the country are concerned this capability could be used to identify and target demonstrators and opposition figures ahead of the 2021 polls. Similarly, in April 2018, Chinese AI firm CloudWalk signed a deal with the government of Zimbabwe to help build a mass facial recognition system. The AI facial recognition system in use in the Ugandan capital is part of Huawei’s ‘’Safe City initiative’’. This technology is already replicated or will soon be replicated in Kenya, Botswana, Mauritius and Zambia. While the deployment of technologies such as these can be useful in curbing crime, they could also become instruments of oppression in the hands of repressive regimes.

Globally, there is also a growing adoption of AI applications for recruitment of human resources, credit scoring and even in criminal justice administration. These critical decision-making roles which were once the exclusive preserve of humans have huge consequences for those affected by their decisions.The greatest concern with the deployment of these systems is the bias inherent in the algorithms underlying the AI, which are usually trained with data which excludes members of a population. This leads to decisions and outcomes which further exacerbates marginalization.

A high profile example were reports in 2019 which suggested that Apple card, a credit card created by Apple and developed by Goldman Sachs, was seemingly biased against women by giving them less favourable credit limits compared to men. Another concern is the opacity surrounding these AI systems. Citing trade secrets or confidentiality of patents, owners of these AI systems are reluctant to share the source code powering these algorithms. Before these systems or their variants become widely adopted in Africa, policies which protect human rights must be in place to protect the vulnerable and marginalized.

To benefit from AI, Africa must shore up the gains of the 2IR and 3IR        

More African countries need to join countries like Kenya and Tunisia in having AI national strategies which drives coordinated national efforts towards AI development. Furthermore, in addition to solidifying the policy landscape around data infrastructure, digital skills and human rights protection, in order for Africa to benefit optimally from advances in AI which is a keystone technology of the Fourth Industrial Revolution (4IR), there must be a recognition that the 4IR is only a continuum of the 2IR and 3IR.

The countries which have benefited the most so far from the 4IR are those which have continuously invested and improved on the foundational infrastructructure underpinning the 2IR and 3IR – stable electricity, efficient mass transportation (e.g. efficient rail systems) and reliable and fast broadband access, amongst others. As of today, electricity supply and Internet access in Africa is non-existent for large segments of the population, or provided inadequately where existent. Without access to basic, foundational infrastructure like excellent power and broadband, Africa’s 4IR development will be stymied. However, with continuous investments in these sectors as well as new technologies of the 4IR such as AI, Africa might as well turn a corner, and begin a new chapter of development the continent has never seen.  

The author, Babatunde Okunoye is a research officer at Paradigm Initiative.