Talk about working smarter: NASA scientists searching for craters on Mars train AI software to do the job for them
- Reference: 1601625427
- News link: https://www.theregister.co.uk/2020/10/02/nasa_mars_crater_ai/
- Source link:
The American space agency's Mars Reconnaissance Orbiter ( [1]MRO ) has been circling, and photographing, the Red Planet for more than 14 years, and by comparing historical and fresh images from the probe, boffins try to discover craters formed by meteorite strikes. The problem is that, once the snaps are transferred back to Earth, it takes someone the best part of an hour to analyse a single picture from the MRO's cameras and identify potential crash sites for closer inspection, whereas an artificial neural network can do the same job in mere seconds.
“AI can't do the kind of skilled analysis a scientist can," [2]said Kiri Wagstaff, a NASA computer scientist, on Thursday. "But tools like this new algorithm can be their assistants. This paves the way for an exciting symbiosis of human and AI 'investigators' working together to accelerate scientific discovery."
[3]
Now you see us ... The crater cluster spotted by the machine-learning software. Credit: NASA/JPL-Caltech/University of Arizona. Click to enlarge
Once a possible impact zone is identified, the orbiter's super-camera – known as the High-Resolution Imaging Science Experiment ( [4]HiRISE ) – is used to check for sure. This equipment has such fine resolution, it can pick out the tracks of NASA's Curiosity rover.
And now the team's AI has identified a cluster of what appears to be small impact craters around 13 feet (4 metres) in diameter that were created sometime between March 2010 and May 2012.
To get this far, this a team of computer scientists at NASA’s Jet Propulsion Laboratory built a classifier to detect craters, and trained the model with 6,830 images; some contained known depressions in the Martian surface, others had no craters in them at all.
Next, the trained model was deployed on a supercomputer to crunch through NASA’s collection of 112,000 images taken from the orbiters cameras. "It wouldn't be possible to process over 112,000 images in a reasonable amount of time without distributing the work across many computers," said agency computer scientist Gary Doran. "The strategy is to split the problem into smaller pieces that can be solved in parallel."
“There are likely many more impacts that we haven't found yet," added Ingrid Daubar, a scientist at NASA and Brown University in the US. "This advance shows you just how much you can do with veteran missions like MRO using modern analysis techniques."
The American agency hopes it can one day perform the image processing onboard spacecraft without having to hand the data over to scientists on Earth first. That, however, requires rad-hardening specialized hardware that can perform inference in space at a reasonable level of power consumption. ®
Get our [5]Tech Resources
[1] https://www.nasa.gov/mission_pages/MRO/mission/index.html
[2] https://www.jpl.nasa.gov/news/news.php?release=2020-188
[3] https://regmedia.co.uk/2020/10/02/crater.jpg
[4] https://mars.nasa.gov/mro/mission/instruments/hirise/
[5] https://whitepapers.theregister.com/
Re: What's that you say, NASA?
Yep, and it isn't "Artificial Intelligence" in any shape or form, there is no intelligence there, just an image parsing algorithm.
Re: What's that you say, NASA?
The modern definition of "Artificial intelligence", stripped of all the hype, is something like this:
We want to write a program to do X (e.g. identify impact craters), but that sounds like it would be hard.
So we'll start with a generic program that takes some input, mushes it all up with a huge list of constant numbers, and produces a result. This is our "model"
Then we'll get some suitable input (photos of Mars), and manually analyse it to figure out what answers we want.
Then we'll pick a bunch of completely-random numbers for the constants, and then repeat that so we have a few thousand different sets. We'll run the program over the input, with each set of constants, and give it a score as to how closely its answers matched what we want. (Probably nowhere near, at this stage). Then we'll choose the sets of constants that gave better results, and mix them together to make a few thousand new bunches of possible constants, in the hope that some of those will give better results. Repeat until the answers are "good enough". (They are unlikely to ever be perfect). This is called "training", and the program + constants we end up with are a "trained model". This is likely to take a long time on powerful computer(s).
Now we have our program! And we're going to call it an "artificial intelligence" because we didn't program it to detect craters, it "learned", so that must make it an "artificial intelligence"!
Don't get me wrong, there's a bunch of really smart people coming up with better models, better ways of mixing constants during training, and faster hardware & software to make all this work. There are also a bunch of smart people who understand this and know how to apply it to a specific problem - the best choices of models and training methods, the best way to score each attempt, the way to make it run fastest, etc.
But under the hood, this is still just "let's use oodles of brute-force to write a program to imperfectly solve this problem". It's nothing to do with the human-like "artificial intelligences" you see in books or movies.
Star Trekkin'
It's life, Jim, but not as we know it"
"It's a crater, Jim, but not that you'd recognise it as such"
"It's life, Jim, but not as you'd know it"
Re: What's that you say, NASA?
Definitely not symbiosis but one of the few times the word synergy could justifiably be used.
Re: What's that you say, NASA?
*Marketing* hype?
The term Man Machine Symbiosis was the title of a 1960 paper by Joseph Carl Robnett Licklider - the man who foresaw and gained funding for ARPANET, and to who the founders of Xerox PARC credit as the father of the GUI. Oh, he was also the project director for a precursor to UNIX. Okay yeah, you can say he 'sold' the concepts to the people holding the purse-strings, but he was a psychologist and engineer, not a mere marketing bod.
So, not a newly coined term, and I assumed that most researchers in the field are familiar with him. Certainly none of them let the biological root of the term Symbiosis cause them unhelpful confusion.
Boot, virus, icon etc etc - all terms borrowed by IT from other fields, without any great confusion.
What's that you say, NASA?
"an exciting symbiosis of human and AI 'investigators'"?
No, it is not. Symbiosis refers to two forms of life living together. It's right there in the word, if you look. Last time I checked, computers, even those running AI systems, were not a form of life no matter how hard you squint at them, and regardless of marketing hype.
All this is is another tool at the disposal of the researchers.