The earthquake in Afghanistan shows the promise and limits of AI

These are scenarios that disaster management-focused artificial intelligence experts are trying to prevent. In recent years, there has been a storm of technology and research trying to help governments better anticipate and respond to disasters such as floods, tsunamis and earthquakes.

Researchers are using deep learning algorithms for it filter and city ​​noise so that earthquake data can be better collected. Algorithms analyze seismic data from previous earthquakes to predict earthquakes earlier and alert people faster.

Artificial intelligence can be very fast and 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 many lives.

Afghans mourn the dead, seek refuge after the devastating earthquake

But the situation in Afghanistan, the researchers point out, shows the structural challenges facing artificial intelligence in countries with devastated infrastructure.

To learn more, The Washington Post spoke with Monique Kuglitsch, chair of a joint AI-focused working group on natural disaster management, and Mousavi.

This interview has been edited for length and clarity.

What has AI got to do with earthquakes?

delightful: For earthquakes, there is real-time forecasting and communication using AI. This means detecting an event using real-time data transmissions and projecting what will happen in the coming days, weeks, months or seasons. There is also anticipation or assistance of communication through an early warning system or decision support system.

Musavi: In recent years, using deep learning and advanced artificial intelligence, we see promising results in predicting and predicting earthquake. The technology can predict the intensity of earthquakes based on what seismic monitoring stations have experienced. You can use those observations to predict within a few seconds what the intensity of the oscillations will be.

In Afghanistan, how could AI have helped?

Musavi: Because AI can be very fast, it can give more warning time to people. Increasing it to 20 seconds, 30 seconds to one minute. The warning can be very helpful in saving lives, especially in cases like Afghanistan, where most of the buildings in the quake zone were poorly constructed one-story buildings. They were not very tall or large. In that situation, even 10 seconds can save a lot of lives because people can save quickly.

The quake in Afghanistan killed more than 1,000 and injured 1,600, officials say

delightful: At best, we would know in advance that an event would happen. Then, when the event happened, we would have activated an early warning system so that people could be evacuated in time. Once the event happened, we would have these kinds of decision support tools, control panels and chat robots that would help communities recover immediately.

What challenges do AI solutions face in countries like Afghanistan?

Musavi: In Afghanistan, there was probably not so much chance that an AI earthquake warning system would help. The main problem is that those early warning systems use signals around 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 nearest seismic station that exists in that region is in Kabul, which is 150 kilometers away from the earthquake.

Afghans are hungry as US and Taliban officials blame each other

delightful: The first challenge will be the availability of data. In Haiti, if I am not mistaken, they have 10 seismic monitoring stations in operation. This is very little for an island hit by some quite devastating seismic events. Of course, you also need to have stable communications, infrastructure and electricity, all of these things. You also need computing power to execute these patterns. Anything we can do to support the development of such an infrastructure will be extremely beneficial for all regions, but especially for those regions that have been hit hardest.

What broader challenges do artificial earthquake intelligence solutions face?

Musavi: The data we use [is mostly] subjective. In terms of earthquakes, it is really hard to say where the earthquake happened, its location, its magnitude. The tools we have for this are simply drawing conclusions using data and some traditional techniques, but you are still just estimating. It is not like a picture of a dog versus a picture of a cat, which is much easier to label.

delightful: There is very limited long-term data on earthquakes. Instrumental data, seismic data, data from instruments go only 150 years ago. And a complete earthquake cycle can last thousands of years. You basically need thousands of years of data to model an earthquake. We do not have thousands of years of data. At best, we can get paleo data, which is when you look at what is happening in sediments and use them to understand how often earthquakes have occurred.

So for that, we turn to things like our physical understanding of earthquakes and seismic patterns. And we do these basically like lab earthquakes. And we use them in AI-based models to basically assess how they should respond. It’s the best we have. But of course, you know, there is nothing that can replace an instrumental measurement.