Small Language Models (SLMs) are seen as a key opportunity for Nigeria and Africa to advance AI innovation, particularly in light of the inaccessibility of Large Language Models (LLMs) due to high resource demands and infrastructure challenges. Experts emphasize SLMs’ adaptability and reduced operational requirements, which can support local businesses and governments in integrating AI more effectively, while also cautioning against their limitations in complex language processing.
Experts suggest that Small Language Models (SLMs) represent a viable pathway for Nigeria and Africa to foster innovation in Artificial Intelligence (AI). Following the launch of ChatGPT in November 2022, Large Language Models (LLMs), including Google’s Gemini and Microsoft’s Co-pilot, have showcased the vast potential of AI across various applications, including generative text and speech.
Nevertheless, as noted by Olubayo Adekanmbi and Ife Adebara, LLMs are often beyond reach for regions like Nigeria due to their massive computational needs and reliance on extensive datasets. In his 2024 remarks, Bosun Tijani, Nigeria’s Minister of Communications, emphasized the necessity for the country to engage actively in the global AI landscape.
While LLMs like GPT-4 can have over 175 billion parameters, SLMs usually consist of tens of millions to under 30 billion parameters. The high energy and computation requirements of LLMs render them impractical in regions lacking robust infrastructure. Recently, Nigeria’s AI strategy highlighted the challenges posed by insufficient digital infrastructure in realizing its ambitions in the AI sector.
The country’s strategy aims to develop localized infrastructure and computational capabilities to support AI advancements. Olivia Shone from Microsoft pointed out that SLMs are tailored for specific tasks and are significantly less resource-intensive, thereby enhancing accessibility and cost-effectiveness.
Both Adekanmbi and Adebara argue that SLMs provide a viable solution for sustainable AI in emerging markets, countering challenges related to limited resources, access to digitized datasets, and budget constraints. They state that SLMs lower the barriers to entry for governments, small businesses, and individuals wishing to adopt generative AI technologies.
SLMs are crucial in bridging the digital divide, making AI accessible to underserved populations, and facilitating digital transformation across various industries. The model’s efficiency and adaptability make it particularly suitable for mobile-driven economies.”
Moreover, the offline usability of SLMs in connectivity-challenged areas ensures inclusion in the application of AI technologies. According to Libing Wang and Tianchong Wang, SLMs enable development tailored to the unique challenges of Global South countries, promising a future where accessibility and efficiency in AI are prioritized.
Although SLMs present numerous advantages, the World Economic Forum cautions that they have limitations, such as a narrower capacity for complex language tasks and reduced performance on intricate queries.
In conclusion, Small Language Models offer a groundbreaking approach to Artificial Intelligence in Nigeria and Africa, addressing critical challenges associated with digital infrastructure and accessibility. By reducing resource requirements and promoting local development, SLMs could catalyze a significant transformation in the region’s AI landscape, adapting to the unique needs of emerging markets. However, it is essential to remain cognizant of their limitations in handling complex language tasks and work towards overcoming these challenges for broader applicability.
Original Source: businessday.ng
