Sarah Parcak
Satellite Archaeologist, University of Alabama at Birmingham
Winner of the TED Prize for 2016. Like a space age Indiana Jones, Sarah uses satellite imagery to locate lost ancient sites. Winner of the prestigious 2016 TED Prize, she is making civilization’s invisible history truly visible and offering new understandings of our past. Her multi-disciplinary team is dedicated to identifying and protecting the world’s cultural heritage one extraordinary site at a time.
During the VMworld Innovators keynote on Thursday morning of the conference we heard from Sarah Parcak, @IndyFromSpace, on how she and her team are finding and protecting archeological sites from looting. She illustrated how she and her team have been able to use various techniques to find new sites where no one had expected them to appear. She then transitioned into how looters have been ransacking sites and destroying world heritage.
To help protect these world treasures she started the site www.globalxplorer.org where, soon, it will be possible for supporters to participate in identifying looting by looking at satellite imagery and flagging looting sites. This is known as crowd sourcing or turking, allowing the power of many to identify areas of concern rapidly.
Essentially, a user will be able to visit www.globalxplorer.org and review images of various sites (it has yet to be released on the site). Each area is a given size and stripped of coordinates (so looters can’t use the technology to either evade or find new sites). Users then review the image and tag it as being either being undisturbed or needing further investigation. Presumably this image would be shown to several individuals to help reduce false positives.
This sounds awesome and we can all agree that we should protect our heritage. Wouldn’t it be great if this program could be expanded and be made even more efficient allowing more users to participate with less interaction and fewer false positives?
I know it sounds absurd, have more participants with fewer interactions. How can this be done?
As some of you may know I am a participant in Berkeley Open Infrastructure for Network Computing (BIONC). I participate in projects such as World Community Grid, Large Hadron Collider, and many others. BIONC is a program / app that I install on my compute systems (yes you can even get it for cell phones). It then takes advantage of spare computing power to find potential solutions. For example I’ve participated in looking for ways to prevent the Zika virus.
Wouldn’t it be great if we could protect archeological sites using BIONC? This could actually be done, and the ground work is being laid for it with Sarah’s crowd sourcing.
You’re probably asking, how can something that is looking to find a way to combat Zika also help prevent looting?
It is possible to take the data gathered in the crowd sourcing phase to teach computers how to look for potential sites on their own. This is done by leveraging deep learning. It is possible to build a distributed system that knows what looting looks like. At which point, images can be processed by a distributed network of computers with minimal human interaction.
In other words, we can know from the crowd sourcing what looting looks like, we can teach a deep learning algorithm to identify the looting, and then distribute that across a grid based computing platform to process small chunks of data on a massive scale. This would speed the identification of looting sites and expand the number of locations that could be protected and preserved for future generations.
It would be very cool to see Sarah and her team work with BIONC, NVIDIA, and others to create a distributed deep learning system that can help protect the world’s heritage. I know it’s a project that my spare compute can go towards.
Until, someone develops such a project on BIONC I strongly encourage everyone to take the time to help protect archeological sites, our shared heritage, from looters. You can do this by signing up for notices which will tell you when you can be part of the crowd sourcing beta program and then take a few minutes to parse some photos.
Thank you Sarah for sharing with us at VMworld.
Images from: http://www.sarahparcak.com/ and http://ancientart.as.ua.edu/tag/looting/ (used in deep learning graphic)
