World Bank urges developing countries to build small AI models

The World Bank is encouraging developing countries to create “small AI” systems tailored to local needs, rather than relying on increasingly sophisticated consumer products from Silicon Valley. These specialised models use narrower datasets for specific tasks and can operate on mobile phones with low connectivity. Paschal Donohoe, the bank’s chief knowledge officer, said locally adapted AI could help compress economic development from centuries into decades.

Donohoe pointed to Indian farmers participating in the bank’s AgriConnect programme who need soil analysis in their own language. With insufficient scientists and agricultural experts to serve every small farm, phone-based tools could help fill information gaps. He suggested governments, private companies and the World Bank could collaborate to develop such products.

At the World Bank and International Monetary Fund annual meetings in Bangkok, IMF managing director Kristalina Georgieva stressed that adoption requires electricity, digital infrastructure, public readiness and protection against bad actors. The World Bank has highlighted electricity and internet access gaps in rural schools in sub-Saharan Africa. Georgieva also warned that 40% of jobs worldwide were likely to be augmented or eliminated over time, with young people particularly exposed as routine entry-level roles disappear.