Steve Chien talks about the @NASA EO1 sat that is controlled using SVM classifiers and Bayesian Thresholding the precursors to AI 
Hyperspectral imaging using superpixel segmentation
Using Bayesian Thresholding on EO1
Using random forest decision trees on Eyafallajokul eruption in 2009
Unsupervised learning using novelty detection (saliency)
Flood and plume detection using saliency detection
Steve is interested in the application od symbolic AI tonretask EO sats
EO1 eventually had to be burned up on reentry when it ran out of fuel
Now the focus has moved to the M2020 rover with improved instruments
Up to 20% uptime for processing can be saved by having advanced scheduling on board
Here are the activities that can be scheduled
AI was also used on the @ESA philae lander
On Mars noone can hear you sneeze and teams of future @NASA rovers will explore caves on Mars that humans will be able to shelter from solar flares
Deep learning is being used to analyse @nasa EO images
@NASA is also doing text analysis to link corpus of papers
The data problem in EO imagery isn't new
@NASA wants to use unsupervised edge AI to build maps and assets on board exploration vehicles rather than sending the back to earth via a low BW link
Europa has more water than earth but to get to it you have to melt through 19km of ice
The ultimate @NASA goal is an all AI mission to our nearest star proxima centauri
