How prepared is Africa for AI?
Introduction
AI is not new. We've been using it for years. Apple's Siri uses machine learning in its speech recognition to learn in order to answer your requests; Google Maps uses an AI powered algorithm to suggest the best routes and transportation for you to get to your destination, and Netflix uses AI algorithms to analyse your viewing history and recommend content based on your preferences. AI has been working behind the scenes for a while, but lately it has become a buzzword. Open AI’s ChatGPT (which has been upgraded to GPT-4) took the world by storm in 2022, with the ability to generate, edit, and iterate with users on creative and technical writing tasks, such as composing songs, writing screenplays, or learning a user’s writing style. The company’s other AI system DALL·E 2 can create realistic images and art from a description in natural language. Around the world AI is being applied in education, health care, agriculture, manufacturing, retail, commerce, governance, and other sectors, allowing more efficient resource allocation which leads to increased productivity and in turn development. In Africa there have been talks of AI, but before African nations attempt to leapfrog their problems they should ask themselves how prepared their country is for AI.
AI in Africa
The African countries that have notable AI solutions are Kenya, Nigeria and South Africa. Kenyan health tech startup Zuri Health for instance provides healthcare services to patients using chatbots that work 24/7 to provide quick responses in real-time to patients and connect them to healthcare products and services. There is no need to call a hospital. Another solution of note is from Zenvus, a Nigerian company that sells a smartfarm hardware that lets farmers know what is happening in their farm in terms of humidity, light, nutrients, ph, and temperature. These factors are not detectable by the human eye. The user sees this data in real time through an app that also gives them suggestions, taking away guesswork. It can further detect disease, drought, pests and stress through a special camera. In South Africa various companies are either integrating AI solutions to existing operations or developing new AI solutions. South Africa in fact leads the continent in AI adoption with a robust ecosystem that includes numerous technology hubs, research groups, and various. AI solutions on the surface seem like leaps in the right direction applicable to the whole continent. Not quite. The continent in general faces many barriers to AI deployment.
AI barriers
Lack of skill sets
AI requires machine learning and natural language processes (NLP). The latter enables computers to comprehend, generate and manipulate human language. These are highly complex algorithms that require top notch programming skills.
It’s easy to evangelise AI as a saviour. Living in Mozambique and travelling recently to Eswatini, I’ve heard educated people in conversation and politicians in interviews saying “AI will help with commercial agriculture”, “AI will facilitate healthcare”, “AI is the future”. The reality is AI is nothing without human resource competency. AI skills are difficult skills to master, and are in high demand in technological hubs worldwide, just to put it into perspective a Machine Learning engineer in California earns on average $145,475 a year. African companies would likely not pay that amount to an engineer, so naturally they would be attracted to a job in a tech company outside of Africa. Generally speaking Africa has a tendency to cut corners, unfortunately for Africa one cannot just manifest technology. It is not just necessary to learn how to code but more importantly how to think like a programmer and how to think outside of the box. Commendably, Kenya has introduced coding into its formal education curriculum. Meanwhile the school curriculum in most other African countries is still based on memorization and regurgitation. Here in Mozambique some children are so disconnected from school that they spend more time learning how to twerk than on anything of professional use.
Lack of AI government policies
The General Data Protection Regulation (GDPR) is a regulation in EU law on data protection and privacy. This is pertinent to AI because it requires oceans of data that shouldn’t breach an individual's private data. In Africa, South Africa, Nigeria, and Kenya have passed some data protection laws. Botswana, Rwanda, Egypt, Mauritius, Tunisia and Zambia have AI strategies to guide adoption, addressing questions of how to train the workforce, and what kind of infrastructure is needed to build local AI solutions. It’s still early days for these policies and strategies though. Experience says that most of the African population, politicians and decision makers not excluded, have a “wait-and-see” attitude towards life, and technology is no exception. The continent will not be able to use AI without a strategic roadmap.
Lack of data
Most people do not understand how AI works. They look at a chatbot like GPT-4 and think that it is autonomous, self-aware and that it will one day “take over”. AI is none of these things. The name Artificial Intelligence is very misleading, in actuality AI is not real intelligence but an advanced pattern recognition system. AI is trained and relies on oceans of data (big data) to provide responses to the user. The data has to be accurate otherwise: garbage in, garbage out. There are so many situations where AI provides inaccurate responses because it is trained on inaccurate data. The system also can fail if the information requested by the user is not in the data bank. Unfortunately, there is a huge deficit of data from Africa. There is not enough empirical research or documentation. Mozambican universities graduate thousands of students annually, but the amount of graduates who conduct research are in the low two digits. Worse, said research is not necessarily accurate or of high quality. The country has an endemic of plagiarism and “half-assery”. The reality is that most of the research done in Africa is by foreign companies, particularly NGOs, this would be the bulk of data that would feed machine learning algorithms that train AI. Consequently, there would be a degree of bias if an AI system is trained on mostly foreign data.
Lack of stable network connectivity
According to the World Bank only about 36% of the African continent had access to the internet in 2021. In general, network connectivity is slow throughout the continent. In Mozambique I use the biggest internet service provider in the country, with a roughly 12 Mbps download speed, and yet it still oscillates throughout the day, everyday! Although GPT-4 works offline, a solution such as Zenvus’ Smartfarm app in Nigeria or Kenya’s Zuri Health does not. Adoption of AI for nationwide life changing solutions requires a stable and reliable internet connection, otherwise it would not be a viable solution.
Language and usability barriers
With the exception of countries such as Seychelles, Mauritius, Tunisia, Kenya, Algeria, Ghana, Egypt, Namibia, Libya and South Africa, illiteracy rates are still very high throughout the continent and public school education is not the best. Children go to school in a language that isn’t their own, it’s merely the official language adopted through colonialism. So much knowledge is not transmitted in a comprehensible way to students through the official school curriculum. In light of this, AI is a very advanced concept to comprehend. User issues relating to language and functionality are inevitable. Zuri Health in Kenya has good usability as it provides digital healthcare in English, Swahili and French. Swahili is widely spoken in Kenya and Tanzania, it is used in education, administration, business, entertainment and other spheres. However, in countries like Mozambique and Angola the traditional languages are not officially used in the system. With the exception of the Bible, and language courses, literature in the traditional languages is practically non-existent. For there to be a good user experience advanced AI solutions should be offered in the user’s dominant language.
The problem of usability harkens back to education. It’s not enough for a solution like Zenvus’ Smartfarm to just exist if no one can use it properly. Users should be able to operate it and perform basic troubleshooting. Most people, anywhere, are not tech savvy and when you factor in educational deficiencies and language barriers it creates a bigger wall. Strategies on training should be carefully thought out before attempting blanket modern solutions such as AI.
Closing thoughts
AI systems are permeating all aspects of life, they are there in the apps and gadgets we use. These are all for generalised purposes, not made to solve nationwide problems. For AI to succeed there has to be a vibrant local ecosystem featuring universities, policy makers, start-ups, large companies, and multi-stakeholder partnerships. While countries like Nigeria, Kenya, Rwanda and South Africa are taking noticeable steps forward and can be considered African tech hubs, this isn’t a reality in most of Africa. People seeing advancing AI in other parts of the world can dream of it as a solution for problems in education, health care, agriculture, manufacturing, retail, commerce, and governance throughout Africa. They should wake up and smell the coffee, AI does not just manifest itself out of thin air, it requires highly competent human resources, funding and industry specific data, big data. Furthermore, the issue of ethics needs to be addressed through standards, procedures, and policies. AI will not come to Africa’s rescue as a handout. Our leaders and policy makers should start by educating the children on programming, and slowly but steadily reform the system and perhaps the next generation will create effective local technological solutions for local problems.
