The Pentagon Is Turning to Nature to Solve Its Most Complex Problems
DARPA is exploring ways to harness chemical reactions, biological processes and other natural phenomena to build a more efficient computer.
Modern computers aren’t powerful enough to handle the Pentagon’s most complex models and simulations, so the department is looking to nature for a new solution.
On Aug. 1, the Defense Advanced Research Projects Agency kicked off a research initiative that looks to harness the computational power of living cells, chemical bonds and other natural processes to develop more efficient computers. Participants in the program, called Nature as Computer or NAC, will rapidly explore a wide array of computational processes found in the natural world and prototype systems that can mimic them in a lab.
In the years ahead, the most promising projects could spin off into their own full-fledged DARPA programs, with the potential transforming aviation, robotics, nanomaterials and an untold number of other fields, according to NAC Program Manager Jiangying Zhou.
“We're trying to learn from the mechanisms nature is using and then engineer materials to mimic that process,” Zhou said in a conversation with Nextgov. “A lot of those [computational] feasibilities are already established by the research community. NAC is pushing us to go one step further, saying, ‘OK, now all these are possible. Can they be used to solve really hard problems?’”
Related: DARPA Thinks Insect Brains Might Hold the Secret to Next-Gen AI
Related: DARPA Is Taking On the Deepfake Problem
Related: DARPA Seeks Space Robots to Repair Satellites
The NAC program seeks to address the same fundamental challenge that’s fueling the government’s quantum computing research: the silicon-based, binary computers that exist today aren’t equipped to solve the complex problems the U.S. faces today. Using classical computers to simulate nuclear detonations, model air turbulence or predict other intricate physical processes consumes a significant amount of time, money and hardware, which even the Pentagon can’t always afford.
“Without a change in computing technology, the level of power consumption, fault tolerance, and cost necessary for large-scale multiphysics modeling may be enormous and impractical,” DARPA officials wrote in the NAC solicitation. While quantum technology offers one solution, research has also shown natural processes could help solve these problems with a fraction of the resources.
Take protein folding, the process by which proteins built by living cells contort themselves into the three-dimensional structures that let them function properly. Protein folding is “exponentially complex” for scientists to model on a computer, but in the real world, the entire process is completed in a matter of milliseconds, Zhou said.
Researchers still don’t fully understand the mechanisms at play in protein folding, “but one of the insights from the current scientific understanding is nature seems to be seeking a different computing strategy,” she said. And instead of using a separate piece of hardware to complete the computation, like a server or mainframe, the process occurs within the protein itself.
“It doesn't need to be plugged into a power outlet … but if you're running the same algorithm on the computer ... it consumes a huge amount of power,” Zhou said. “That's where the efficiency is coming from.”
By learning how to harness this process, she said the Pentagon could potentially build airplane wings that automatically adjust their shape for different conditions or so-called “soft robots” that can adapt to their environment through decentralized intelligence, much like an octopus.
Besides protein folding, there are numerous other natural processes that could potentially yield the same sort of efficiency; researchers have found light waves, chemical reactions, and certain composite materials all have properties that humans could potentially co-opt to solve computationally complex problems.
Under NAC, participants will select a natural computational process they wish to exploit and complete a proof-of-concept demonstration by the end of the 18-month program. DARPA will offer each team up to $1 million in funding.
The program is hosted under DARPA’s Disruptioneering initiative, which offers small bursts of funding for high-risk, high-reward research efforts. The agency adopted a similar model to support its more ambitious artificial intelligence projects.
If NAC results in a prototype or idea that proves to be “at least theoretically feasible,” DARPA intends to launch a bigger, better-funded program explore it further, Zhou said.
Teams interested in participating in NAC must submit their proposals by Sept. 3.