Small Language Models: Nigeria’s Pathway to AI Innovation

Small Language Models (SLMs) are seen as a promising avenue for AI innovation in Nigeria and Africa, offering practical solutions amid limited infrastructure. Unlike Large Language Models, SLMs require fewer resources, making them more accessible and cost-effective for local needs. These models can drive digital transformation across various sectors, particularly benefiting mobile-driven economies and underserved communities.

Small Language Models (SLMs) are poised to present Nigeria and Africa with significant opportunities for advancing Artificial Intelligence (AI) innovation, according to expert analyses. The introduction of ChatGPT in November 2022 has revealed the potential of Large Language Models (LLMs), leading to the development of models such as Google’s Gemini and Microsoft’s Co-pilot, which aim to expand the capabilities of generative text, speech, images, and videos.

However, SLMs are considered more feasible for Nigeria where LLMs often remain inaccessible due to the required extensive computational infrastructure and large datasets. In a recent discussion, Olubayo Adekanmbi, founder of Data Science Nigeria, and Ife Adebara, an AI expert, pointed out that while LLMs like GPT-4 might possess over 175 billion parameters, SLMs operate effectively within the range of tens of millions to under 30 billion parameters.

The notable energy and computational requirements of LLMs hinder their practicality in regions with underdeveloped infrastructures. In releasing its AI strategy draft, Nigeria acknowledged that insufficient digital infrastructure is a barrier to its ambition of being a leader in AI development on the continent. The strategy aims to cultivate affordable and localized infrastructure along with the necessary computational capacity to facilitate AI growth.

According to Olivia Shone, senior director of product marketing at Microsoft, SLMs focus on specific, less resource-heavy AI tasks, thereby enhancing accessibility and cost-effectiveness. “SLMs can respond to the same queries as LLMs, sometimes with deeper expertise for domain-specific tasks and at a much lower latency.”

Adekanmbi and Adebara, co-founders of EqualyzAI, highlight that SLMs provide a viable pathway for sustainable AI growth in emerging markets. The advantages of SLMs include minimized computational requirements, lower inference latency, cost efficiency, and ease of customization. Furthermore, they significantly lower the entry barriers for governments, businesses, and individuals seeking to incorporate generative AI into operations.

“SLMs therefore represent a revolutionary approach to bridging the digital divide and making AI accessible to those who need it most,” the founders asserted. They argue that SLMs can accelerate digital transformation across sectors in Nigeria by catering to its mobile-driven economy and enabling offline access for underserved rural communities.

Experts Libing Wang of UNESCO and Tianchong Wang from Swinburne University emphasize, “SLMs hold immense promise for shaping the future of AI, particularly in scenarios where accessibility, efficiency, and affordability are critical.” Nonetheless, it is essential to note the limitations of SLMs as identified by the World Economic Forum, including their reduced accuracy in complex tasks and narrow performance scope.

In summary, Small Language Models (SLMs) present Nigeria and Africa with an indispensable opportunity for fostering AI innovation. They provide a practical alternative to the excessively resource-demanding Large Language Models (LLMs), ensuring accessibility and efficiency in regions with limited infrastructure. As various experts underscore the potential of SLMs to transform local economies and drive technological advancements, a balanced perspective on their capabilities and constraints remains crucial for future developments in the field of AI.

Original Source: businessday.ng