While movies may give us the impression that being a spy is an endless parade of gun battles, covert infiltrations, and drinks, it can actually be quite a boring job.
According to Robert Cardillo, head of the National Geospatial-Intelligence Agency, “a significant chunk of the time, I will send [my employees] to a dark room to look at TV monitors to do national security essential work. But boy is it inefficient.” With the perhaps hundreds of thousands of photos and videos captured every day, it’s hard to argue with Cardillo. Assigning humans to comb through the data requires them to look at each piece of collected intelligence individually, which is a long and drawn out process requiring a huge number of employees. But, Cardillo hopes to address this inefficiency with the recent advances in artificial intelligence.
Artificial intelligence’s ability to analyze vast amounts of image and video data allows it to greatly outperform its human counterparts. Algorithms can find patterns, recognize landscapes, and identify unusual objects much quicker than its human counterparts. This would allow AI to spot foreign missile-silo activity or the movement of troops on a country’s borders faster than a human spy could, which in turn would influence the speed in which our defense agencies react. AI can also monitor and analyze multiple sources simultaneously, meaning a single computer server could perhaps replace hundreds of human spies.
To achieve his goals, Cardillo has hired former Findyr CEO Anthony Vinci to lead the development of the National Geospatial-Intelligence Agency’s machine-learning group. Cardillo’s initiative isn’t the first use of AI by the intelligence community though. DARPA and IARPA, the US defense and intelligence research agencies, have funded deep-learning (the basis of AI) since the 1960’s, and the federal government spent a whopping $1 billion on AI research in 2015 alone.
Even with this investment though, the US government is still trailing Silicon Valley in the development of AI driven tools. Google’s AI platforms can recognize items in satellite imagery so well that they can be used in aviation and defense platforms, while Facebook’s AI training module can recognize 40,000 images a second. With Cardillo and the National Geospatial-Intelligence Agency hiring of Anthony Vinci, perhaps we’ll see a bit of Silicon Valley’s efficiency make its way to the defense and intelligence community.
Images: Wikimedia, Pixaby, Eon Productions
Source: Foreign Policy