You've probably never looked at a mammal’s brain and thought “Gee, I wish I could yank that out of its skull and shrink it onto a chip.” Nor have you likely gazed upon a colony of ants and remarked “wouldn’t it be great if we could get spy drones to work together like that?”
That’s because you don’t work for the Defense Advanced Projects Research Agency, the Pentagon’s way, way out science and technology arm. Their annual budget request, which they made public on Thursday, reads like something out of lost a Philip K. Dick notepad.
DARPA, for the uninitiated, acts as the Pentagon’s blue-sky research agency, always looking beyond the horizon for the technologies which will have the greatest impact in the future. If you’ve ever found yourself wondering, Joker-style, where the U.S. military gets those wonderful toys—like the Internet, global positioning systems and stealth bombers—chances are it started out as idea on a drawing board at DARPA.
This year the Obama administration requests nearly $3 billion DARPA for the research outfit—a nearly $136 million increase over the agency’s last budget year. Tucked away inside that $3 billion are a number of new and fascinating projects: ones to make faster, more cooperative unmanned systems, to mimic parts of the human body for smarter computers, and to even build prosthetics that feel like real hands.
For instance, the $10 million Human and Computer Symbiosis program will teach computers to recognize when it encounters a bit of question that only a trained, flesh-and-blood expert would know—and then ask one of us meatbags for the answer. The computers will shoot a text to a predefined list of experts, learning more about the subject over time. Eventually, the plan is for the computers to become experts themselves and able to provide answers when asked a question.
DARPA’s Cortical Processor, however, takes the human-machine interface a step further by looking to mimic the mammalian neocortex. As it turns out, the cortex in mammals’ brains is pretty darn good at processing large amounts of data in real time and controlling multiple motor functions. Computing power like that can come in handy. So the Cortical Processor program will spend $2.3 million trying to develop a chip that behaves like a neocortex and equip it with a series of algorithms known as Hierarchical Temporal Memory (which are themselves based on the neocortex) to effectively create a cortex on a chip. The resulting chips could be used in battlefield systems such as unmanned aerial vehicles to more quickly make sense of the gobs of data hovered up by the military’s various surveillance sensors. The Cortical Processor builds off the research from a previous DARPA project “SyNAPSE,” which sought to make a chip which could imitate the function of a cat’s cortex. The program managed to produce a chip with “1 million neurons performing behavioral tests in the virtual environment,” according to DARPA.
You've probably never looked at a mammal’s brain and thought “Gee, I wish I could yank that out of its skull and shrink it onto a chip.” That’s because you don’t work for the Defense Advanced Projects Research Agency.
But DARPA doesn’t just want machines to get smarter; it wants them to work better, together. We’re talking about drones, which have already changed how the U.S. wages war. But the potential of drones is far from untapped and this year’s DARPA budget is brimming with research for new ways to leverage advances in robotics. The $5 million “Swarm Challenge” looks to see if a flock of small drones can all play well together. It involves the development of algorithms that would allow a number of small unmanned systems to work in unison and solve problems. Darpa envisions the drone hive-mind could be useful supporting troops in air, ground and maritime operations and could even help out in obstacle-clearing and search and rescue operations.
Drones in the skies get most of the attention these days, but they’re far from the only flavor of unmanned systems the U.S. military is interested in. Undersea unmanned vehicles have been getting increased attention from defense researchers these days but they come with a handful of technological hurdles their airborne colleagues lack. Among them, getting a vehicle to move fast underwater—and developing the energy systems to support it—is particularly tricky. So DARPA’s turning to the action movie-titled “Blue Wolf” program to develop technology that can create a super-fast underwater drone. At a price tag of nearly $14 million, the program looks to leverage the lessons learned from an earlier project, the “underwater express,” which reduces the water’s drag on a vehicle by surrounding it with a bubble of air.
Not all of DARPA’s work is focused on making new gadgets for far-off battlefields. Some of the most important projects in this year’s budget submission have to do with taking care of wounded warriors right here at home. Towards that end, the Prosthetic Hand Proprioception & Touch Interfaces program (shortened as HaPTIx) aims to make better prosthetics for troops and veterans with amputated limbs. Here DARPA is looking to spend $7 million create nerve implants which could offer amputees not just greater motor control and sensory awareness of their prosthetics. It’s a tall order, one involving the development of new nerve interface technologies as well as new surgical techniques. If it’s successful, though, it could provide an important advances in prosthetics for both military and civilian amputees alike.
Of course, just because these projects exist on paper, doesn’t mean they’ll pan out in practice. Part of DARPA’s raison d’etre is to work at the edges of possibility to see what works—and what doesn’t. And feasible or not, it’s a fiscally-constrained Congress which has the final say on funding levels and whether these projects will end up as science fiction or science fact.