Google, Harvard Use AI To Predict Earthquake Aftershocks

Scientists from Harvard and Google have fed a neural network with historical seismological data "and more accurately predicted where 'more than 30,000 mainshock-aftershock pairs' from an independent dataset occurred, more accurately than previous ways like the Coulomb forecast method," reports Engadget. "That's because the AI method takes multiple aspects of stress shifts into account versus Coulomb's singular approach." From the report: This model isn't ready for primetime yet, though. The scientists note that their study only counts one type of aftershock triggering when making predictions (static stress changes), rather than accounting for static and dynamic stress changes. "The combination of static and dynamic stress changes leads to a spatial distribution of aftershocks that differs from the pattern caused by static stress changes alone," Nature writes. Then there's the fact that the models don't take complex faults into account when making predictions. The science journal explains this shortcoming as such: "This could explain why the authors see no evidence of a lack of aftershocks near faults -- caused by an overall decrease in stress -- despite the fact that this feature is readily apparent in situations in which data and circumstances allow it to be clearly observed."

Google ads?

[Anonymous Coward]

This is so google will know when / where to display mobile insurance ads?

Re:

[grep -v '.*' *]

This is so google will know when / where to display mobile insurance ads?

No, you misunderstand -- you sell insurance to people that WON'T use it. If you sell to people that use it and over-saturate (not enough unaffected people to cover the affected ones), you're screwed. You hopefully have one outgoing payment hit and a whole lotta misses, which translate to a whole lotta income that you keep or have a whole bunch of hits and go broke. The whole point of insurance (from the insurance company's standpoint) is that the customer will pay for it and then NOT use it.

OTOH if you

Re: Google ads?

[Zero__Kelvin]

I think it is you that misunderstood. If you know when and where earthquakes will happen you also know when and where you can sell insurance to people who will never need it.

Algorithm on computer != AI

[Excelcia]

Running a predictive algorithm on a computer is not AI. Please, everyone, stop using the term AI for everything.

"More accurate prediction..."

[hcs_$reboot]

I was wondering how can they can validate precisely the new model, given the - fortunately - few events (large earthquakes) occurring a year. The explanation comes from one of TFAs, "The authors withheld a randomly selected 25% of the mainshock–aftershock sequences from the training data, and used this subset to validate the predictive power of their machine-learning method". So they used the input that built the neural network to validate the model... that, validates the neural network algorithm, but

Re:

[religionofpeas]

So they used the input that built the neural network to validate the model

No, they used 75% of the data as input to build the model, and then used 25% as validation test.

Re:

[hcs_$reboot]

It seems we have a different view concerning the meaning of that "withheld"