Small Language Models (SLMs) are being identified as a key factor in promoting AI innovation in Nigeria and Africa. Experts argue that SLMs are more accessible and feasible than Large Language Models (LLMs), which require significant computational resources. Nigeria’s AI strategy emphasizes the need for localized infrastructure to support AI development, and SLMs are highlighted as a sustainable solution. Despite some limitations, SLMs are viewed as a means of facilitating technology advancements in under-resourced regions.
Experts suggest that Small Language Models (SLMs) could catalyze Artificial Intelligence (AI) innovation within Nigeria and the broader African continent. The advent of ChatGPT in November 2022 has showcased the potential of Large Language Models (LLMs) such as Google’s Gemini and Microsoft’s Co-pilot, which have inspired diverse applications in generative text, speech, images, and videos. However, SLMs present a more accessible alternative that can drive AI advancements in regions with limited infrastructure.
Olubayo Adekanmbi, founder of Data Science Nigeria, and AI expert Ife Adebara assert in their white paper that while LLMs like GPT-4 are robust, requiring substantial computational infrastructure and extensive datasets, SLMs—with parameters ranging from tens of millions to under 30 billion—are more suitable for Nigeria’s infrastructural constraints. They emphasize that the disparity in accessibility could hinder Nigeria’s ambitions to be an AI leader in Africa.
The Nigerian government recognizes the challenge posed by inadequate digital infrastructure in its draft AI strategy, which lays out plans for developing affordable and localized infrastructure capable of supporting AI. Olivia Shone, senior director of product marketing at Microsoft, notes that SLMs focus on specific tasks that are less resource-demanding, enhancing their practicality. “SLMs can respond to the same queries as LLMs, sometimes with deeper expertise for domain-specific tasks…but they can be less accurate with broad queries,” Shone pointed out.
Adekanmbi and Adebara describe SLMs as a sustainable solution to AI development challenges in emerging markets with limited resources and digitization. They emphasize that SLMs lower barriers for governments, small businesses, and individuals aiming to leverage generative AI in their operations. By requiring less computational power, SLMs can effectively address the digital divide and facilitate the integration of AI across various sectors.
The unique capabilities of SLMs make them well-suited for mobile-driven economies like Nigeria, where offline operation is crucial due to inconsistent internet access. Experts Libing Wang and Tianchong Wang advocate for SLMs as a vital tool for developing AI applications that reflect local expertise and challenges.
While SLMs offer significant advantages, they do come with limitations such as reduced capacity for complex language and constrained performance on intricate tasks. Nonetheless, these models hold considerable promise for enhancing accessibility and affordability in AI development, particularly in underserved regions.
In conclusion, Small Language Models (SLMs) present a viable pathway for Nigeria and Africa towards embracing AI innovation. They address critical challenges posed by limited infrastructure while enabling affordable access to AI technologies. Though SLMs come with their own limitations, their potential to drive local development and support mobile economies positions them as a transformative force in the digital landscape.
Original Source: businessday.ng
