Why Representative Data Is Critical to Nigeria’s AI Ambitions

In the pursuit of Nigeria's AI ambitions, the importance of high-quality, locally relevant data cannot be overstated. In my previous work this year, I emphasized that AI flourishes due to the availability of data on the internet, a point I now consider as the most critical fact often overlooked in discussions about Africa's role in technology.

The ability of AI to function across diverse domains – language, vision, medicine, logistics, among others – is rooted in the vast datasets it is trained on. These datasets encompass the public web, licensed sources, code, books, images, audio, video, and both human-generated and synthetically created material. The scale of these data sets lays the foundation for all subsequent advancements.

One exemplary model is AlexNet, a cornerstone in image recognition, which was made possible through the use of 1.2 million labeled images. Though this represents a fraction of the data today's models utilize, it nonetheless revolutionized the field. The core lesson isn't that you need web-scale data to matter; rather, it's that you need sufficient, high-quality data. However, for Africa, including Nigeria, this is currently lacking.

This gap is particularly evident in the medical field, where AI's promise of personalized medicine – tailored treatment plans, drug recommendations, and risk models based on an individual's genetics, history, and context – faces significant challenges. For many African populations, realizing this potential is hindered by the underrepresentation of African genomic data, disease patterns, and clinical outcomes in the datasets used to build and validate precision-medicine tools.

While the underlying science holds, models trained on datasets insufficiently representative of African demographics are less capable of generating reliable, locally relevant recommendations.

Nigeria's infrastructure challenges, such as power stability, compute access, and high connectivity costs, are indeed real. Yet, the country can still position itself favorably for the AI age by focusing on the data it can collect and generate locally. A prime example is Nigeria's national demographics. The country's last official population census was carried out in 2006.

In contrast, countries like South Africa have led in HIV research and resource allocation due to their robust testing and digital tracking protocols, which provide clear, actionable data. Conversely, Nigeria's reliance on paper-based record-keeping across many institutions creates significant gaps in understanding national health trends.

Data collection directly impacts the daily lives of people and can also aid in combating counterfeit products. For instance, an AI system designed to identify fake medicines, drinks, or other consumer goods would require verified examples of genuine and counterfeit products, including packaging images, batch numbers, barcodes, registration details, and expert validation. Such a system could assist regulators and consumers but should still be overseen by humans and not considered absolute proof of authenticity.

Nigeria's challenge thus lies not only in "building AI" but in constructing the data systems necessary to enable useful and contextually relevant AI. This involves more than mere digitization; it requires reliable data collection, interoperable systems, secure storage, privacy protection, skilled personnel, and clear rules governing data access and use.

Crucially, data from Nigerian communities should be actively collected, owned, interpreted, and benefit those who contribute it. This means integrating local languages, cultures, markets, health conditions, and social realities into the datasets used to develop future AI systems. Without this, Nigeria risks being a mere consumer of technologies that do not fully comprehend its people.

To truly position Nigeria for AI, a national data agenda must take center stage. This agenda should focus on digitizing essential records, enhancing data quality, supporting research, connecting public institutions, safeguarding citizens' privacy, and making responsibly anonymized data available for legitimate innovation. Remember, while data is a foundational element of AI, it is not the sole determinant of success.

The true potential of AI in Nigeria will be realized only when reliable, representative, and responsibly managed data becomes the bedrock upon which intelligent systems are built.