The commercialization of space and the high volume of visual information now being collected has revolutionized geospatial intelligence. This means agencies like the National Geospatial-Intelligence Agency and the National Reconnaissance Office must find new ways to innovate, according to Jack O’Connor, a retired CIA and NGA executive who is now with the Krieger School at Johns Hopkins University.
Work is being done on computer algorithms that can interpret imagery similar to an analyst, and that is becoming increasingly important.
“The volume of information is such that it has made a different scarcity. Years ago, when space satellites were new, what was scarce was images. but now it’s the attention of the analysts that’s scarce,” O’Connor said on Agency in Focus: Intelligence Community.
The NGA and NRO are focusing their efforts on gathering what visual information they can uniquely and buying the rest from commercial vendors. O’Connor says simply buying digital images is not enough, the agencies need to be able to quantify how valuable what they are purchasing will be for them.
“So the concept of interpretability used to be measured and then, for some reason about 10 years ago, they stopped measuring it. But there are other units of measure that would help socialize the concepts that NGA and the NRA are trying to do and also help the commercial vendors,” O’Connor said.
Some imagery can look good but lack value while others may not appear as technically sound but hold great value. Having common scales and terminology in the community would go a long way for innovation.
Despite all the innovation that has led up to now, many of the challenges of the past remain true. Especially human capital.
“Previous director of NGA Robert Cordero estimated that it would take more than a million analysts to look at what was coming down, and that was a few years ago. So there’s even more imagery being created now, and there are not government resource is for those kinds of investments in that number,” O’Connor said.
AI will play a key role in filling those gaps, but it is not the one-stop solution. The algorithms must also be held accountable for errors just like a person would be. This requires a sound auditing process with a clear unit of measure so issues can be identified and improved upon.
A combination of humans and smart algorithms is a balance all interested parties are chasing.