Over the past decade, generative artificial intelligence has become a powerful tool for creating jobs for tertiary-educated youth in Africa, but challenges of digital work on the continent remain.
However, according to the African Union Development Agency – New Partnership for Africa’s Development (AUDA-NEPAD), the continent’s primary technical and development body, many African tertiary institutions offer artificial intelligence (AI) courses that have outdated curricula and graduates who are not ready for digital work.
In a position paper, ‘AI and the Future of Work in Africa’, AUDA-NEPAD stated that governments and universities have not invested appropriately in cutting-edge technology, and the education content is highly theoretical. The paper was prepared jointly with the University of Oxford, Microsoft and the Centre for the Future of Work at the University of Pretoria in South Africa, among other stakeholders.
Highlighting what Africa should do to achieve a breakthrough in dignified digital work, the paper also warned that, although generative artificial intelligence has sparked a global conversation about its benefits and impact on the future of work, not everyone in Africa is suited for such work.
According to the panel that wrote the paper, even at its peak, generative AI may not be equally useful for everyone, and its impact might not be evenly distributed across regions and communities on the continent.
In short, generative AI is the use of AI to create new content, such as text, images, music, audio and videos.
Training data dilemmas
“But its success will depend on the amount and quality of the training data and how decisions about design, languages to be used, and training are made,” wrote the panel that included Dr Barbara Glover, a programme manager at AUDA-NEPAD, Jacki O’Neill, the director of the Nairobi-based Microsoft Research Africa, and Dr Vukosi Marivate, the chair of data science and machine learning at the University of Pretoria.
In that context, AUDA-NEPAD stressed that the continent’s demographic and socio-economic realities should be considered for AI to impact Africa’s future work. “For now, most training data for existing generative AI models is sourced from the English-speaking Global North,” noted the paper.
So far, the current stream of generative AI models uses limited training data in African languages. For instance, only Kiswahili, Afrikaans, Kinyarwanda, and Igbo were selected in the training data for the large language model GPT-3. Other African languages used in the AI training series include Amharic and isiZulu.
To date, fewer than 10 African languages have been used in AI training models on a continent with more than 1,000 languages, adversely affecting AI workers who would like to use their own languages.
“Besides, the amount of training data from African sources used by the current crop of AI models is also limited, which means that African contexts are likely to be underrepresented in those models,” noted the paper.
Data cost
Even assuming that generative AI has the potential to transform African economies by providing jobs to the expanding youth bulge, AUDA-NEPAD and its associates explained that there are profound implications in that most of the continent’s population is not exposed to AI.
“To be able to take advantage of even the off-the-shelf AI systems requires access to devices and data, knowledge of their existence and the skills to make use of them, as well as the ability to translate use into desired outcomes,’’ according to the report.
The issue is that only half of African countries currently have computer skills in their school curriculums, compared to a global average of 85%. As a result, an estimated 95% of Africa’s youth aged 15-24 work in informal economic settings, making a living as street vendors, taxi drivers, hairdressers, metal workers, and in repair shops.
In this regard, the International Monetary Fund, in a working paper published early this year, predicted that generative AI is expected to have a delayed impact on African economies.
The problem of AI exclusion in Africa is expected to continue – even in countries with low barriers to access to generative tools and platforms, as many workers lack access to data.
“Outside of North Africa, African consumers pay disproportionately higher prices for mobile phone data, with some of the most expensive countries worldwide being in Sub-Saharan Africa,” noted the panel.
The cost of data is the highest in Sudan, Uganda and the Republic of the Congo, and the lowest in Kenya, Nigeria and Ethiopia.
Exploitation of digital workers
Concerns about biases in AI algorithms, the scarcity of quality African data, toxic language and questionable labour practices in the digital work sector have raised red flags about the much-talked-about future of many well-paid digital jobs on the continent.
One of the most notable cases that have been reported on focused on the treatment of Kenyan digital workers involved in data collection for training AI models two years ago.
According to an article in Time, OpenAI, the creator of ChatGPT, outsourced the work to Kenyan digital workers to make the chatbot less toxic and paid them meagre wages.
The data labellers employed on behalf of OpenAI were paid a take-home wage of between US$1.32 and US$2 per hour, depending on seniority and performance, stated Time.
These workers were recruited by Samasource Impact Sourcing, Inc, formerly Sama, the United States firm that provides training in big data and AI, which had established a sweatshop in Nairobi where digital workers from various African countries were paid low wages to moderate Facebook content.
“Despite their importance to Facebook, the workers in this Nairobi office were among the lowest-paid workers for the platform anywhere in the world, with some of them taking home as little as US$1.50 per hour,” stated Time.
What is emerging is that Nairobi – and probably other African cities such as Addis Ababa, Johannesburg, Lagos and Accra – are likely to become hubs for outsourced content jobs.
The issue is that English-speaking digital workers would make it easy for companies outside the continent to set up satellite offices in those cities and take advantage of tertiary-educated young people desperate to get jobs.
In effect, there are indicators that many digital workers, referred to in Kenya as ‘digital hustlers’ or merely AI labourers, face exploitative working conditions and lack protection.
Beyond those working in content moderation and labelling, many university graduates in Nairobi are working on digital platforms providing services in health care, beauty, taxi, delivery, and personal service and selling in e-commerce.
Digital workers are poorly paid
According to a recent International Labour Organization (ILO) report on digital labour in Kenya, other digital jobs in Nairobi are in tutoring, which is a misleading description for those providing services on academic cheating contract platforms. Almost all the workers in digital tutoring in Nairobi are graduates or university students.
But what is shared among digital labourers in Nairobi is that they are poorly paid. The ILO says digital workers in Kenya earn US$2 per hour compared with the average of US$12 to US$20 per hour earned by their counterparts in the Global North for the same work.
The US$2 per hour earnings was an increase in response to a published news article in Time magazine before when, according to the workers, they earned about US$1 per hour, noted the ILO report published in March 2024.
Subsequently, based on the experience of digital workers in Nairobi, AI work is not the dignified labour that it is often imagined to be.
As AUDA-NEPAD has pointed out, given the widespread exploitative and deplorable working conditions of low-wage data annotators across the continent, it is evident that the AI sector is likely to replicate existing extractive and often unethical labour practices in Africa.
AUDA-NEPAD, therefore, recommends that African countries must design policies and regulations that strengthen the development of AI while limiting its adverse effects.
About The Author
Related posts:
Nigeria: Students under 18 are too young for Higher Education, minister decides
BANGLADESH: Universities closed as protests against quota system escalate
USA: Could the US benefit from the international student backlash elsewhere?
Morocco: Anger as dean ‘confronts’ student with Palestinian keffiyeh
Missing Brit Ben Ross found ALIVE in Majorca after ‘series of unfortunate events’ led to him vanishing 12 days ago
Perezida Kagame yihanganishije Abaturage ba Ethiopia bibasiwe n’ibiza
Kigali: Abaturage 12 bajyanwe mu bitaro nyuma yo kunywa ubushera
Chris Brown n’itsinda rye bari mu mazi abira
