While the IC’s research organization looks into adding security to cloud environments, in the here and now, intelligence agencies are sharing more data.
Picture a disaster in a remote part of Africa.
Following news of the event, the U.S. military is dispatched to an area where residents speak a little-known language with multiple dialects. En route to the site, the emergency team feeds the location and the region’s primary linguistics into a database that deciphers enough conversation to enable military commanders to communicate with local leaders.
Such a capability would offer U.S. officials a chance at triage before a humanitarian aid effort begins and, in the process, improve recovery efforts, says John Launchbury, the former director of the Information Innovation Office at the Defense Advanced Research Projects Agency (DARPA). Pentagon leaders now say this kind of machine learning and data analysis is critical to maintaining military effectiveness.
Machine learning is a type of artificial intelligence that allows technology to evolve as it absorbs new information. With superfast processing, networks can ingest huge amounts of data — from logistical movements to stealth fighter activities — and apply it to improve decision-making.
“Machine learning can help our operators sift through these ‘data haystacks’ to find the needles that need their attention,” says Army Lt. Col. Roger Cabiness II, a DOD spokesperson.
The Army Research Laboratory considers AI and machine learning “game-changers” for future warfighting capabilities, says Brian Sadler, an ARL senior scientist in intelligent systems. “Artificial intelligence and machine learning are now embedded across many areas of research.”
The technology could predict supply shortages in logistics systems; anticipate equipment failures for maintenance; update tactical platforms to respond to operator habits; and allow sensors to recognize targets, Cabiness says. “The possibilities are endless.”
DOD is deploying speech-recognition technology on aircraft flying over contested areas to simultaneously track hundreds of conversations on radio waves that discuss a targeted individual or terrorist activity, Launchbury says. Humans can track only a half-dozen channels at once.
But learning machines do more than accelerate and streamline tasks, especially those beyond human capabilities, says Denise Zheng, senior fellow and director of the Technology Policy Program of the Center for Strategic and International Studies. They can analyze vast amounts of data in seconds, adding statistical rigor.
Zheng points to DARPA’s 2016 Cyber Grand Challenge. There, seven computers sought loopholes in one another’s code. One found a bug in another, infiltrated it and extracted data. Meanwhile, another identified the same weakness in itself and developed a patch in less than 20 minutes. “This is using AI to help fix vulnerabilities in software at machine speed,” Launchbury says.