On: AI accurately predicted 70% of earthquakes a week ...

[hyperhopper]

~61% is what that sounds like.

Even lower it you take the f1 score...

[gpm → hyperhopper]

I think it sounds more like 93% (14 predicted/(14 + 1) total earthquakes).

False positives aren't failures to predict an earthquake, they're a different error with a different consequences.

Of course, I'm not sure why I would trust the absolute numbers reported are more accurate than the percentages reported...

[bhaney → gpm]

Sounds good. In that case, I just made a 100% accurate model by always predicting that every location on Earth is experiencing an earthquake all the time.