Machine Learning For Earthquakes
Machine learning is getting as good at predicting earthquakes as we are at predicting that Beetlejuice will be showing up on streaming services again this month.
What happened: Three recent studies showed how incorporating machine learning models helps outperform current state-of-the-art models for forecasting earthquake aftershocks.
- Earthquake forecasts are not weather forecasts. No seismologist on your local news channel will be warning of a magnitude 3.0 in the Bay Area tonight at 5 pm.
Instead, these forecasts are built to understand trends, particularly those after main shocks rather than telling us when the next big one will happen.
How do the models work? Large Language Models (LLMs) like ChatGBT seem smart, but really they’re just statistics trained on millions of words. Similarly, these earthquake models have been trained on large datasets collected by seismologists that then use machine learning to improve the performance of estimating aftershocks.
Why it matters: Industries continue to figure out ways to incorporate artificial intelligence and machine learning to improve their performance.
- In the summer, a group of researchers found incorporating an LLM into autonomous driving systems helped it better predict human behaviour beyond the simple sensor system.
Zoom out: Whether it comes to improving earthquake forecasts or speeding up autonomous driving models, we’re just beginning to see the impact of machine learning and AI in industry.
(Our next big forecast: Mariah Carey start ticking up on Spotify rankings in about a month.)