Anticipating food crises: AI and food stability

Food crisis experts are harnessing artificial intelligence to tackle global food instability. A new study has found that deep learning can anticipate food crises, which gives agencies the opportunity to respond with the appropriate humanitarian aid.

The study was conducted by the World Bank and leveraged deep learning technology by inputting over 11 million news articles focused on food-insecure countries. By sifting through the data, the AI was able to generate an early warning system which outperformed existing food crisis risk models.

One of the authors of the study, Samuel Fraiberger, explained that traditional risk models miss the bigger picture, so they often fail to predict imminent food crises. "Historically, when you look at the data, the forecasts made by these experts, they are quite imprecise," commented Mr Fraiberger.

For example, traditional food crisis models concentrate on quantitative data, which is often incomplete — particularly in data-scarce areas — or fails to capture the bigger picture. Instead, machine learning can harness a mixture of qualitative and quantitative data to generate a holistic understanding of real-world scenarios which better anticipate food crises.

The study uses the example of a food crisis in Ethiopia in 2009 which was not anticipated by the traditional model because the precipitation index failed to capture the actual effects of a drought. In contrast to the traditional model, machine learning was able to retrospectively flag the drought as a risk factor by interpreting news articles.

Using AI and machine learning opens up a world of data which can be modelled to anticipate potential disasters. Ultimately, this buys disaster response agencies time to allocate humanitarian aid. These discoveries are timely: as climate change grinds down on our food systems, technology that supports food stability is more important than ever.