Hi! 👋 Welcome to The Big Y!
When you live in California, conversations about the timing of the next big earthquake are inevitable. Using AI we’ve improved significantly on natural disaster and weather forecasting but given the unpredictable and varying nature of earthquakes it is more challenging than something like hurricane or flood tracking. But just because something is challenging, doesn’t mean that researchers aren’t working to apply novel AI approaches to earthquake monitoring and detection. With an earthquake, even a couple of seconds of notice can make a huge difference.
Out of Stanford, researchers have built QuakeFlow, which uses a deep convolutional neural network combined with another ML algorithm, GaMMA, to identify earthquakes along with the approximate locations and magnitudes. While not necessarily a forecasting tool, QuakeFlow does help with increasing earthquake data catalogs with scalable data transformation, which then creates a foundation for forecasting models.
Researchers at the Los Alamos National Laboratory have taken Meta’s speech recognition algorithm, Wav2Vec-2.0, and applied it to seismic signals coming from a set of moderate magnitude earthquakes in Hawaii. Adapting from speech waveforms to seismic waveforms, the model aimed at near-future prediction of earthquakes. The results aren’t quite there yet, but it does open the door to potential predictability.
Tools like earthquake detection and prediction have the potential to save lives and it is interesting to watch the algorithms made for chatbots expand into new and more impactful areas.
The Tidbit: It turns out that using AI tools for performance reviews is actually helpful and especially beneficial for women who tend to downplay their achievements. Having a more standard and uniform performance review tool helps level the playing field and give more objective outcomes.
Know someone who might enjoy this newsletter? Share it with them and help spread the word!
Thanks for reading! Have a great week! 😁