Targeting the Islamic State, or Why the Military Should Invest in Artificial Intelligence
Bottom Line: Integrating artificial intelligence (AI) into surveillance activity will allow military unites to hit a greater number of hostile targets with increased accuracy and less uncertainty.
A lot of preparation goes into identifying hostile targets and preparing a strike
When Hans Vreeland's unit got word that ISIL was storing in weapons in one of a hundred bunkers scattered in an area the size of 13 Central Parks, it couldn't simply level them all. First, the unit had to determine precisely which bunkers were being used to house the weapons.
This is a tedious and time-consuming task
To determine where to strike, a pilot in Vreeland's unit flew a drone in circles, capturing hours of video of a single spot on the property. Every day, Vreeland reviewed the footage and summaries of activity, and occasionally selected images to be pulled from the feed. This continued until enough information was gathered, assembled into a PowerPoint, and presented to headquarters, who approved the strike. This process was repeated in some form or another before every strike, sucking up a lot of the time and energy from Vreeland's small unit.
AI can hasten this process and make units more efficient
Vreeland suggests two specific technological improvements. First, the use of autonomous drones programmed to "search specified areas and identify activity," would free up time for the pilots who would otherwise be flying the drones in circles. Second, the adoption of something like Project Maven, a machine-learning program that reviews videos and still images and identifies objects of interest, will cut down on the time people like Vreeland will spend reviewing footage. Of course, AI won't replace the human element, so military personnel will still be making the final judgment calls.
Integrating AI does not mean sacrificing precision
Vreeland is clear about the risks that come with this type of targeting: it's not always easy to tell the difference between four men with guns and four men with fishing poles. Ultimately, it comes down to men like Vreeland to decide whether or not they should hit a target. But tools like Project Maven could be helpful when it comes time to make these "moral decisions." For instance, a human may not be able to tell the difference between a gun and a fishing pole in a drone image, but a computer might. The precision of AI helps minimize uncertainty in crucial moments.
AI isn't completely ready for battle -- but it will never be ready, unless the military uses it
Vreeland doesn't claim that these tools are still being developed, and likely still have unforeseen flaws. But, he argues, these flaws will only become apparent once these technologies are used in the field. Additionally, Vreeland believes that successfully integrating these technologies in the field will encourage further investment in AI, thus further refining them, and providing additional benefit to military units.
Read the full article here.