By the end of 2017, the Pentagon hopes that computers, not human soldiers or analysts, will be spearheading the hunt for Daesh militants and hideouts in Syria and Iraq by using aerial surveillance footage and turning it into "actionable intelligence."
Visiting robotics lab at University of #Illinois pic.twitter.com/GKg8w2md7p
— Robert Work (@DepSecDefWork) May 14, 2017
This is part of Project Maven, an effort launched last month by Deputy Defense Secretary Work to engineer new types of machine learning in the US Defense Department.
The memorandum for Project Maven lays out quite clearly its raison d'etat: "as numbers studies have made clear, the department of defense must integrate artificial intelligence and machine learning more effectively across operations to maintain advantages over increasingly capable advisories and competitors."
Reportedly, military and civilian analysts are "overwhelmed" by the sheer amount of video footage that they have to sit down and sift through to look out for behavior indicative of Daesh terrorists. The footage is taken by drones and coalition aircraft flying over Iraq and Syria in the fight against Daesh in both countries.
While there is plenty of footage, and much of it recorded on high-resolution camera, analysts are finding it difficult to deal with the vast amount of intelligence that they have to work with. Leaders in the Pentagon are hoping that changes in technology will be able to life some of the burden off of the stretched workforce, while simultaneously bringing better results on the battle against Daesh.
Currently, analysts have to go through the laborious task of filling in any abnormal behavior they spot on the footage into a spread sheet. In an attempt to to increase efficiency, last month, the Pentagon established the Algorithmic Warfare Cross Functional Team. The team will be developing artificial intelligence to sort through the massive amounts of video data collected by human surveillance drones.
Essentially, the purpose of the unit is to reconcile the huge amount of data with machine learning in the military. The team will be working with the Pentagon's Strategic Capabilities Office, which is a group that tries to reengineer and enhance current weapons technology to a more lethal caliber. The hope is that by marrying the two departments, the DoD will be able to hasten the process of automating intelligence for military purposes for effectively.
Work is well known in military circles for emphasizing the importance of focusing on AI. AI was thrown to the forefront of US military concern when it became the fulcrum of the "Third Offset Strategy," which was the Pentagon's initiative under the Obama administration to offset Russian and Chinese technological advancements.
Looking fwd to discussing offset strategy 2mrw at 8am #CSISLive #defense https://t.co/DV1XET5WP0
— Robert Work (@DepSecDefWork) October 27, 2016
Rather than killing machines, Work's idea of AI has been to focus on a collaboration between humans and artificial intelligence. In a nutshell, the machine would sift through massive amounts of data at high speeds while the human intelligence analyst makes the final decisions.
As it stands, the Algorithmic Warfare cell is waiting on a US$70 million approval of funds from the congress. Once approved, the cell will follow a "three-phase effort," laid out by Work, each of which will last a duration of 90 days. The first phase will reportedly be to decide upon data labeling algorithms, the second will be to acquire the hardware or software that can process the algorithms, and the third will be to put the technology to use in military missions.
However, as it stands, AI is still in the basic phases of development. The Pentagon is finding that computers are superior at generating the data, but still lack the ability to analyze it to the extent of humans. This means that for now, the intelligence analysts at Arlington might have to go on being overwhelmed with cell phone intercepts and drone video footage, for quite some time yet.
