Predicting Natural Disasters

When the National Guard wants to see the scale of a natural disaster, it flies planes over a site, then analyzes the images. What once took 12 hours now takes minutes thanks to Bellwether, a project incubated under Alphabet’s moonshot factory, X. Bellwether, which calls itself “the first prediction engine for Earth and everything on it,” pulls insights from images using AI. The system analyzes 600 different layers of geodata, including vegetation levels, wind speed, and precipitation, in its machine learning models. The models include 20 years of data, and run thousands of "what if" questions to try and figure out the chances of a flood, fire, or other natural disaster striking. But it’s more than that. “Before you get to sexy AI, you first need to organize all of this Earth observation data,” says Sarah Russell, who founded Bellwether in 2019. Alongside the government, insurance companies are using Bellwether to better understand the scale of damage caused after a disaster, and how much to insure areas for in the future. Russell says Bellwether may spin out from X within a year.

Learn More at X Bellwether