IBM wants to predict earthquakes and volcanoes with Watson

No more guessing.
No more guessing.
Image: AP Photo/Ben Curtis
We may earn a commission from links on this page.

We may soon have categorical evidence that living in San Francisco is a terrible idea.

IBM announced on Nov. 20 that it had created an award-winning simulation of the Earth’s tectonic plates that could soon be used to make predictions about when the next great earthquakes and volcanic eruptions will occur. And its artificial intelligence system, Watson, may prove to be the computer brain that can tell us when it’s time to get out of the Bay Area.

A team of computer scientists at IBM, in partnership with researchers from the University of Texas at Austin, New York University and the California Institute for Technology, created a model that simulated the entire flow of mantle under the Earth’s surface. The model is so complex that it had to run on the Sequoia supercomputer (the third-fastest computer in the world), which the company built for the US government. The team’s model last week won the Gordon Bell Prize, the top award in supercomputing. 

Costas Bekas, an IBM researcher on the team, told Quartz that while we understand how the Earth’s crust moves, our fundamental knowledge of what’s going on under the surface is still quite limited. “We much more about outer space than the interior of our planet,” he said. That makes it hard to predict and prepare for disasters caused by tectonic shifts—be they earthquakes, volcanoes, or tsunamis. “How can you predict a phenomenon that you don’t understand?” Bekas said.

The ever-moving molten mess below the Earth’s surface does not act uniformly, nor are the gaps between plates regularly spaced. This makes the mathematics “very, very difficult,” Bekas said. The research team took readings from seismic sensors across the planet, and paired them with the pervading geological theories for tectonic movements. Using the data, they created what they believe to be far more accurate mathematical models for the Earth’s interior and ran those models on the Sequoia supercomputer.

Were you somehow able to run this model on a regular home computer, it’d take you three years to get a sense of what’s going on under the Earth’s surface. The Sequoia can do it in a day.

The result does not yet predict when natural disasters are likely to happen, but it does potentially give geologists and seismologists a tool for creating such a system in the near future, Bekas said. Now that there’s a model in place, the next steps will be to feed it more information to improve it.

The increasing proliferation of internet-connected sensors will provide more seismic data points in the coming years. Watson has the ability to find patterns in massive data sets, as well as analyze human writing. If you take every geological research paper ever written, throw in every data point from every seismograph in the world, give them to Watson, and combine those with advanced simulations such as the one performed on Sequoia, the result, Bekas said, would be a system that “will profoundly increase our understanding of the geological processes that drive natural disasters.” And the more data points there are, the more accurately Watson should be able to understand big tectonic movements.

IBM is working with the US Department of Energy on this model, as well as its work on attempting to predict the weather. The company also recently bought most of the assets of Weather Company—the owners of the Weather Channel—including its myriad weather sensors around the world. IBM previously told Quartz that the acquisition “will coalesce around cloud, Internet of Things, and Watson.”

Why does IBM care about being able to predict the weather, let alone when the next earthquake is happening? Because it is looking for impressive AI feats to show off to perspective clients of Watson, on which it has already bet about $1 billion. Being able to show multinational corporations that it knows if Mauna Loa is about to blow might well help it win their data analytics business. That could truly result in some explosive growth.