These are scenarios that artificial intelligence experts, focused on disaster management, are trying to avoid. In recent years, there has been a surge in technology and research to help governments better predict and respond to disasters such as floods, tsunamis and earthquakes.
Researchers use deep-learning algorithms to filter out urban noise so that earthquake data can be better collected. Algorithms analyze seismic data from past earthquakes to predict earthquakes earlier and notify people faster.
“AI can be really fast — it can give people more warning time,” said Mostafa Mousavi, an artificial intelligence researcher at Stanford University who specializes in geophysics and earthquakes. “Even ten seconds can save a lot of lives.”
But the situation in Afghanistan, researchers note, demonstrates the structural challenges artificial intelligence faces in places with crumbling infrastructure.
For more information, The Washington Post spoke with Monique Kuglitsch, chair of a joint UN working group focused on AI for natural disaster management, and Mousavi.
This interview has been edited for length and clarity.
What does AI have to do with earthquakes?
Kuglitsch: For earthquakes, there is real-time prediction and communication using AI. That means detecting an event using real-time data streams and projecting what will happen in the coming days, weeks, months or seasons. There is also forecasting or assistance with communication through an early warning system or decision support system.
Mousavi: In more recent years, using deep learning and advanced artificial intelligence, we are seeing promising results in predicting and predicting ground shaking. Technology can predict the intensity of ground shaking based on what seismic monitoring stations have experienced. You can use those observations to predict the intensity of the shaking in a matter of seconds.
How could AI have helped in Afghanistan?
Mousavi: Since AI can be very fast, it can give people more warning time. Increase to 20 seconds, 30 seconds to a minute. The warning can be very helpful in saving lives, especially in cases like Afghanistan, where most of the buildings in the earthquake zone were poorly constructed single-storey buildings. They weren’t very long or huge. In that situation, even 10 seconds can save a lot of lives, because people can quickly escape.
Kuglitsch: In the best case scenario, we know in advance that an event is going to take place. Then when the event happens, we would have activated an early warning system so that people can evacuate in time. Once the event takes place, we would have these kinds of decision support tools, dashboards and chatbots that allow communities to recover immediately.
What challenges do AI solutions face in countries like Afghanistan?
Mousavi: In Afghanistan, there may not have been such a great opportunity for an AI earthquake early warning system to help. The biggest problem is that those early warning systems use signals in the area, so it depends on how many seismic stations or sensors we have near the earthquake. In Afghanistan, in terms of data collection, it shows that the closest seismic station in that region is in Kabul, which is 150 kilometers away from the earthquake.
Kuglitsch: The first challenge will be data availability. In Haiti, if I’m not mistaken, they have 10 seismic monitoring stations. This is very little for an island hit by some pretty devastating seismic events. Of course you also need stable communication, infrastructure and electricity – all these things. You also need computing power to run these models. Anything we can do to support the development of such infrastructure will be hugely beneficial to all regions, but especially to those hardest hit.
What broader challenges do earthquake AI solutions face?
Mousavi: The data we use [is mostly] subjective. In terms of earthquakes, it’s really hard to say where the earthquake happened, the location, the magnitude. The tools we have for that are just inferring using data and some traditional techniques – but you’re still just estimating. It’s not like a dog photo versus a cat photo, which is much easier to label.
Kuglitsch: There is very limited long-term data on earthquakes. Instrumental data, seismic data, instrument data only go back 150 years. And a full earthquake cycle can take thousands of years. You basically need thousands of years of data to model an earthquake. We don’t have data from thousands of years. At best, we could get paleo data, which is when you look at what happens in sediments and use that to figure out how often earthquakes happened.
So for that we turn to things like our physical understanding of earthquakes and seismic patterns. And we actually make these as lab lab quakes. And we use those in AI-based models to basically estimate how they should respond. It’s the best we have. But sure, you know, there’s nothing that can replace an instrumental measurement.