Back to the Analog Future
September 6, 2012
Wired informs us, “Darpa has Seen the Future of Computing. . . and It’s Analog.” We thought the future was biological (our vision is computers which are alive) but we guess analog is okay.
Through its research arm Darpa, the Department of Defense aims to backtrack from digital processors in their effort to boost power efficiency. It is the roadblock in power scaling, they say, which has recently hampered the progress proscribed by Moore’s Law. Specifically, the Department has been running up against power limitations in drone- and helicopter-mounted spy cameras.
The write up tells us that Darpa program manager Daniel Hammerstrom is:
“. . . looking for an alternative to straight-up boolean logic, where the voltage in a chip’s transistor represents a zero or a one. Hammerstrom wants chipmakers to build analog processors that can do probabilistic math without forcing transistors into an absolute one-or-zero state, a technique that burns energy. . . .
“Analog computers were used in the 1950s, but they were overshadowed by the transistor and the amazing computing capabilities that digital processors pumped out over the past half-century, according to Ben Vigoda, the general manager of the Analog Devices Lyric Labs group. ‘The people who are just retiring from university right now can remember programming analog computers in college,’ says Vigoda. ‘It’s been a long time since we really questioned the paradigm that we’re using.'”
Why yes, it has. The back-to-the-drawing-board approach is often prudent, but it can also be a big risk. Can Darpa break free from the world of ones and zeros? Their program, called UPSIDE (Unconventional Processing of Signals for Intelligent Data Exploitation), has been allotted 54 months, and will have two phases. First, chips will be built using probabilistic techniques. Next, those chips will be used to build mobile imaging systems.
Reporter Robert McMillan notes that, should Darpa be successful, the new probabilistic chips could help with some of the complex data analysis problems researchers have been struggling with. That would be a nice side effect. Will the analog approach pan out? Only time will tell—give them about 54 months.
Cynthia Murrell, September 06, 2012