Scientists from Harvard University found another use for artificial intelligence: to amplify the sensitivity of Oklahoma’s seismographs #machinemagic
Did you really think we wouldn’t get here?! Here, as in to the point where artificial intelligence can improve earthquake detectors and, some hope, predict earthquakes some day. The base is here already – scientists are experimenting with machine learning software to detect earthquakes happening in Oklahoma faster and better than traditional machines.
Why there? While the state is not known for natural disasters, it did register in 2015 more than 900 small earthquakes. Since then, their number has decreased but it’s still in the hundreds.
Authorities are not prepared to detect and locate all these phenomena so they’ve employed artificial help. The one used works similarly to the voice detection software smart assistants rely on. Both are filtering out “noise” to pinpoint specific sounds. In the case of earthquakes, it’s about distinguishing between ambient seismic noise and earthquakes, no matter how small or faraway they are.
The neural network was trained with data from more “quiet” areas; knowing how that ambient noise is, it can bypass that and focus on the needed data. In their experiments, published in the journal Science Advances, the team says the AI detected 17 times more earthquakes than normal seismographs in a fraction of that time. Give it a bit of time, a lot of tweaks and data and it could one day predict earthquakes based on the frequency of earthquakes.