World Bank says 'small AI' could transform developing countries

World Bank officials are promoting "small AI": compact, purpose-built machine learning models that run on smartphones with limited connectivity and handle local languages and datasets. Paschal Donohoe argues these models could compress what might have taken centuries of economic development into decades by delivering tailored services - health, agriculture, education - at scale. He cited AgriConnect, a program helping Indian farmers get soil analysis in their own language, as an example of how narrow, locally trained models can fill expertise gaps that physical specialists cannot. The World Bank envisions developing these tools through partnerships among the bank, governments and private firms to ensure they meet on-the-ground needs.

Speakers at the World Bank-IMF meetings framed small AI within broader infrastructure and labour challenges. IMF managing director Kristalina Georgieva said success requires reliable electricity, digital connectivity, adoption readiness and government protections, warning that AI could augment or eliminate about 40% of jobs and hit young entrants hardest. The bank’s own report highlights major gaps - many rural sub-Saharan schools lack power or steady internet - while the institution has been mobilising private capital ($120bn last year) to support development. The argument is that with basic services and strategic finance in place, small AI could be as democratising as the cellphone.