AI TECHNOLOGY
Just after sunrise on Saturday, a satellite turned its lens towards Mandalay, Myanmar’s second largest city, following the powerful 7.7 magnitude earthquake that struck the day before.
The aim was to capture images, with the help of AI, that could quickly show which areas were worst hit, pinpointing collapsed buildings and helping aid teams figure out where to go first.
But the tech hit an early roadblock.
Thick cloud cover made it impossible to get clear visuals.
“There’s no way to see through clouds with this technology,” explained Microsoft’s Chief Data Scientist, Juan Lavista Ferres.
Hours later, clearer images were captured by another satellite, this one from San Francisco-based Planet Labs.
These were sent to Microsoft’s AI for Good Lab in Washington, where a team worked through the night to analyse the data.
It’s not the first time the lab stepped in during a crisis.
They’ve previously mapped flood damage in Libya and tracked wildfires in LA.
But earthquakes are trickier.
Unlike fires, which spend in more predictable ways, earthquakes affect entire cities in uneven patterns.
That meant a standard AI model wouldn’t do the job; they had to build one tailored specifically for the region affected.
The final analysis showed:
515 buildings were nearly or totally destroyed
1,524 buildings suffered moderate to severe damage
A delay in the sky
Microsoft shared this with relief groups like the Red Cross to help speed up the response on the ground.
The company was clear: this is just a starting point and will still need to be verified by teams on-site.
Since the quake, Planet Labs has taken satellite images of around a dozen locations across Myanmar and neighbouring Thailand to support ongoing efforts.
We built AI for disasters, but nature still humbles us.