ANNA ZAMAN: AI can be a game-changer for Africa

The world stands at the precipice of a huge technological revolution: artificial intelligence (AI). With tools like ChatGPT that respond to a wide range of questions from a human user, and Murf AI — which converts text to speech — the world has been taken by storm. Within the initial week of its launch ChatGPT was able to attract 1-million users.

Such transformative technology showcases the potential of AI and indicates that generative AI is the future of technology. With various industries exploring innovative AI applications in areas such as automotive vehicles, healthcare services and digital education, the transformative power of AI is increasingly evident.

But what does this mean for Africa? How could such a transformative technology work in the modernisation of Africa and its industries? Will it prove to be transformative or dangerous for the population?

There has already been an increase in machine learning presence in Africa, as Google has opened an AI research centre in Ghana that will focus on machine learning research to address the specific challenges faced by the African continent. Other than this, the African Master’s in Machine Learning (AMMI) programme was launched in 2018.

This pan-African initiative aims to train a new generation of African machine learning experts. The program is run by the African Institute for Mathematical Sciences (AIMS) and has partnerships with major tech companies such as Google, Facebook and Microsoft.

Such initiatives all point towards a positive future for Africa as AI holds the potential to drive economic growth and development across the continent. However, it is important to understand that this groundbreaking technology doesn’t come without its risks and possible dangerous implications in Africa.

Machine learning is a part of AI that requires developers to provide data and examples to the computer program. The program learns from these examples and creates algorithms that identify patterns within these data sets to make appropriate decisions about any new data.

However, the training data can be subject to biases and prejudices since it is generated and recorded by human beings who may hold similar beliefs and values. The point of concern for Africa remains within the representation bias that is evident in machine learning.

Representation bias occurs when the training data does not accurately reflect the reality it is meant to model, resulting in over-or underrepresentation of certain groups. If a biased sample of the population is used by a computer to make inferences, it may lead to systematic disadvantages for those who are either over-represented or underrepresented in the data set.

The bias becomes even more evident since these AI tools have been created mostly by dominant white groups who do not accurately represent African individuals. There have been cases of representation biases within AI tools already, including the Amazon hiring case, where the automated algorithm discriminated against women, or the failure of facial recognition technology to capture and recognise the faces of black individuals or individuals belonging to minority groups.

If these discriminatory cases were so evident within the Western world, it is difficult to imagine the extent of the problems and discriminatory practices they would cause in Africa given the additional challenges posed by cultural and language differences.

Though the potential for AI technology to boost economic development in Africa is significant, ensuring that these tools make impartial and defensible decisions remains a critical challenge. Addressing these issues requires a grassroots approach, including improving education around machine learning, incentivising the creation of unbiased data sets or the recording of data with minimal pre-existing biases, and emphasising the importance of localised data processing to ensure AI tools are guided by diverse populations.

AI in Africa must be developed by taking into account the needs and perspectives of local communities. African countries need to prioritise these measures to build a sustainable and inclusive ecosystem for AI development.

Though programmes like the AMMI can help graduate students become proficient in machine learning, it is essential for countries to increase their investment in Stem (Science, Technology, Engineering, Maths) education from an early age. It is essential to establish a more robust curriculum that incorporates computer science and machine learning, with a particular emphasis on advancing overall education development.

If these steps are not taken in a timely manner Africa will be left behind in the AI transformation age. Furthermore, to concretise the presence of Stem-based education in Africa and ensure it is not simply a surface level endeavour by governments, those provided with skills such as machine learning must have an efficient platform to reap tangible benefits.

For example, Andela is a skill-based hiring platform where users can specify what computer-based training they have previously received and get redirected to jobs that require a similar skill set. This way Africans would be further incentivised to pick up skills like machine learning as it would ensure employability and access to a job-market otherwise closed off to them.

In essence, it is critical that the emergence of technology in Africa is a well-rounded effort, starting from grass roots level education to the introduction of a profitable job market. Moreover, once a system is established AI and machine learning tools must be redefined to overcome their inherent Eurocentric biases to increase their accessibility and utilisation.

• Zaman is an independent researcher in Georgetown University's science, technology and international affairs programme, focused on emerging technologies and development.