[mebazaa]

"The AI accurately predicted 70% of earthquakes a week in advance, with 14 forecasts coming true… However, it issued eight false warnings and missed one earthquake." Beyond the fact that n is very small, I don’t think 14/23 is 70%…

[two_handfuls]

My guess:

There were 20 real earthquakes and the AI issued 14 warnings. 20*70/100 = 14.

Of course the problem with that is that it’s always a balance between precision and recall (or specificity vs sensitivity).

Here they show 70% recall which sounds good, but the precision is 14/(14+8)=64% which is decidedly less impressive.

Those terms are defined in Wikipedia here: https://en.m.wikipedia.org/wiki/Precision_and_recall

[0cf8612b2e1e]

Everyone knows that AI gets at least three mulligans.

[jeffbee]

"A week in advance" is overstating it. It forecast the week of the earthquake, counting a hit in a radius of 200 miles. That is probably exciting for the researchers, but of no current practical value.

[spajus]

70% of the time it works all the time