The earthquake in Turkey and Syria in February 2023 took an awful toll on human beings, causing huge loss of lives, properties and indescribable misery.
While preventing natural disasters is a different conversation, minimizing loss of lives, properties, and human suffering is something that we may have some control over — with artificial intelligence (AI) the latest ally in this mission.
The deep learning and neural network characteristics of AI can study historical and current data — everything from tectonic movements, rise in water levels, and water temperature in the ocean — and help warn of occurrences of natural disasters like floods, earthquakes, and volcanoes.
It’s not a one-size-fits-all solution, but it is a tool in the toolkit that teams are already implementing.
How Does AI Predict Natural Disasters?
Let’s take the example of an AI system developed by Google and Harvard to predict earthquakes.
The main challenge to preventing the loss of lives is to identify the location of an earthquake. Aftershocks follow every major earthquake and can continue for a long time, toppling structures already weakened by the parent earthquake and causing more injuries and deaths.
While human experts can predict the occurrences and location to an extent, there is a scope to improve the accuracy and timing of the predictions, which is where AI can come in.
The AI system developed by Google and Harvard analyzes more than 131,000 “mainshock-aftershock” events in a database to understand and spot patterns.
Scientists then tested the neural network on a database of 30,000 pairs of mainshock and aftershocks.
The neural network performed better than existing systems of earthquake prediction, which focus on what is known as the Coulomb failure stress change.
On a scale of 1, where 1 represents 100% accuracy, the Coulomb failure stress change rated 0.583 for predictive strength, while the AI system rated 0.849.
According to Brendon Meade, a professor of Earth and planetary sciences at Harvard:
“There are three things you want to know about earthquakes: when they are going to occur, how big they’re going to be, and where they’re going to be.
Before this work, we had empirical laws for when they would occur and how big they were going to be, and now we’re working the third leg, where they might occur.”
Problems in the Adoption of AI
Using a more everyday occurrence, weather prediction usually falls under government-supported agencies, and yet AI has not really entered the weather station to any large degree.
According to Dale Durran, a professor of Atmospheric Sciences at the University of Washington, “The most innovative work on the modeling itself seems to come from private companies right now more than the [government] weather services. The weather services need to maybe be paying more attention to this. They have a lot invested in the current approach, and it works pretty well, but it’s very computationally intensive.”
Perhaps AI needs to be more widely used, more extensively tested, and possibly cheaper to implement (considering changing from a current system brings its own costs) before it becomes more widely used as a predictive tool.
Reliability is also an important factor in treating AI as a serious tool, and it also has not yet had a widescale ‘success’ in predicting a natural disaster.
Either way, new tools can save lives. For example, in the 1980s, weather models could predict natural disasters three days in advance — and now it has increased to seven.
Progress is being made, but until the tools are put to work, we don’t have any evidence to help decide what role AI can play in predicting natural disasters.
However, the private sector is picking up the tools and pushing them forward, and with companies like Google involved, it is an industry that is likely to move fast — much like everything else in AI.
There are few goals greater or noble in life than saving lives, so anything that moves the dial in this direction will be a good move, and we will keep exploring.