The World Bank is urging developing countries to build “small AI” systems that can run on mobile phones, work with limited connectivity and respond to local languages and conditions. These specialised models use narrower datasets for defined tasks rather than attempting to match the scale of frontier consumer systems.
Paschal Donohoe, the bank’s chief knowledge officer and a former Irish finance minister, said adapting AI to local needs could allow economic development that might otherwise take centuries to happen within decades. He pointed to Indian farmers involved in the bank’s AgriConnect programme who need soil analysis in their own language, noting that there are not enough scientists and agricultural experts to provide that service to every small farm.
The bank’s approach is part of a broader effort under President Ajay Banga to attract private investment into emerging economies. Donohoe said the institution had mobilised $120bn in private capital over the past year through guarantees and other financial instruments, as the United States, the United Kingdom and other countries cut aid budgets. The bank has also described small, low-cost AI as a way to extend services in areas such as agriculture, education and health without requiring large data centres, while warning that limited power, connectivity, skills and institutional capacity can restrict access to AI’s benefits.