Predicting the weather to protect health: seventy years of progress that has never reached poor countries

A University of Chicago report supported by the Rockefeller Foundation believes that AI models can finally make forecasting affordable — as long as they don't replicate existing inequalities.

Published on September 22 by the Institute for Climate and Sustainable Growth of the University of Chicago with the support of the Rockefeller Foundation, the report starts from a simple economic observation: a ten-day forecast once demanded a supercomputer at a hundred million dollars; learning models produce it today in a few minutes on a single chip. The localized forecast of low- and middle-income countries, the very countries that are at risk of the most severe health risks of climate change, can be brought to bear.

Someone needs to use it. The report notes that 81 per cent of national meteorological services provide health-oriented climate services, but only 23 per cent of ministries of health use meteorological data. The main author, Amir Jina, sums up the issue: AI models offer the opportunity to democratize forecasting, so that seventy years of progress will eventually benefit everyone.

Four priorities are being advanced: assessing the forecasts against the health decisions they actually allow, requiring evidence based on health outcomes, creating a mechanism dedicated to national deployment, and including health in national AI policies. Without this, the text warns, the technology will reproduce the gaps it could fill.