Machine learning system can predict slow slip events to within about a week or so. Can't get exact location or magnitude, but for the first time, they can actually predict an event before it happens. To establish whether this worked, they created their own little earthquakes in the laboratory. And that was very much more cessful. So that shows you can successfully transfer from learning on one type of quake, in this case, e ditial simulation, to the real world one. That's exactly what they want to do now, going forward, with real world earthquakes.
Democrats will spend the week battling for a tightening of laws on casting votes; that will
overshadow Republicans’ worrying push into how those votes are counted and certified. Earthquakes remain damnably unpredictable, but
new research suggests a route to early-warning systems. And why hammams, the declining bathhouses of the Arab world, will
cling on despite even the challenge of covid-19. For full access to print, digital and audio editions of
The Economist, subscribe here
www.economist.com/intelligenceoffer