Small Language Models: Paving Nigeria’s Path to AI Innovation
Experts assert that Small Language Models (SLMs) can drive AI innovation in Nigeria and Africa, presenting a practical and accessible alternative to Large Language Models (LLMs). Though LLMs are powerful, SLMs offer reduced computational demands and foster local development. They are especially suited for mobile-driven economies, ensuring accessibility to rural areas and supporting the digital transformation of various sectors in the region.
Small Language Models (SLMs) present a significant opportunity for Nigeria and Africa to advance in Artificial Intelligence (AI) innovation. Experts assert that since the introduction of ChatGPT in November 2022, numerous Large Language Models (LLMs) have showcased the expansive potential of AI, leading to innovations such as Google’s Gemini and Microsoft’s Co-Pilot. These LLMs, however, demand extensive computational resources and vast datasets, making them difficult to access in Nigeria, as pointed out by Olubayo Adekanmbi and Ife Adebara in a recent white paper.
In 2024, Bosun Tijani, the minister of communications, innovation, and digital economy, envisioned Nigeria’s involvement in the global landscape of AI development and regulation. The World Economic Forum reports that while LLMs can have over 175 billion parameters, SLMs typically consist of tens of millions to under 30 billion parameters, making them more feasible for regions with limited infrastructure. The Nigerian draft AI strategy acknowledges that insufficient digital infrastructure could hinder its ambition to emerge as a leader in AI on the continent.
The strategy aims to establish affordable and localized infrastructures, enhancing the computational capacity necessary for AI development. Olivia Shone, senior director of product marketing at Microsoft, highlighted that SLMs target specific tasks and are less resource-intensive, thus being more accessible. “SLMs can respond to the same queries as LLMs, sometimes with deeper expertise for domain-specific tasks and at a much lower latency,” Shone noted.
Adekanmbi and Adebara, founders of EqualyzAI, advocate for SLMs as a sustainable solution for AI in emerging markets. They maintain that SLMs facilitate efficient development, adaptability, and lower entry barriers for governments and small businesses. “SLMs therefore represent a revolutionary approach to bridging the digital divide and making AI accessible to those who need it most,” they said, emphasizing the potential of SLMs to enhance digital transformation across various sectors.
SLMs are beneficial for mobile-driven economies like Nigeria, as they operate effectively with minimal computational power and can function offline, thereby reaching underserved rural areas. Libing Wang from UNESCO and Tianchong Wang from Swinburne University affirm that SLMs can address the challenges faced by Global South countries in digital infrastructure development. “SLMs hold immense promise for shaping the future of AI,” they argued, particularly when accessibility is vital
Despite the advantages, the World Economic Forum has indicated that SLMs face limitations in handling complex language tasks and may not perform as accurately as LLMs in various scenarios.
In light of the discussion on Small Language Models (SLMs), it is evident that these models present a viable pathway for Nigeria and other African nations to engage with Artificial Intelligence. By emphasizing accessibility, reduced computational demands, and adaptability, SLMs can mitigate the infrastructural challenges that have hindered broader AI adoption in such regions. The insights shared by industry experts highlight the role of SLMs not only in technological innovation but also in fostering inclusive growth across diverse sectors.
Original Source: businessday.ng
