After silicon: Biocomputing at work

The warm, fuzzy predictability of Moore’s Law — which stipulates that the number of transistors on a chip doubles every two years — should hold true for another 15 or 20 years. At that point, chip makers looking to downsize transistors will hit a wall. They’ll be churning out silicon using a 16-nanometer process (today’s state-of-the-art chips use a 90-nanometer process). Scrunch things down much smaller and those skittering electrons will be unable to reliably register the 0s and 1s that binary processes require. Semiconductor designers will respond by experimenting with new substances or by stringing together processors in grids and other arrangements. But some breakthrough will be required to replace the dead end that is silicon.

In all likelihood, that breakthrough will be modeled after biological processes. InfoWorld’s own Jon Udell insists he’ll be satisfied by future computers that simulate intelligence by “patterning their behavior on ours.” Many researchers will want to go one better. DNA computing, neural networks, and other efforts architected around biological organisms will offer computing styles uniquely adapted to solving problems that computers currently can’t: interpreting environmental data (think face recognition), performing massively parallel operations, or healing and regenerating themselves. After decades of designing ever denser hardware in an attempt to outmuscle the human mind, the late 2020s will herald a return to organics.

These scenarios aren’t science fiction. Even today, IBM, Hewlett-Packard, and others are pushing self-healing or autonomic computing — initiatives clearly inspired by biological models. Furthermore, the mammalian brain is no longer a black box. Every day we’re gaining further insight into its workings. A group of North Carolina scientists, for instance, recently implanted a device in monkeys that allows the monkeys to move a robotic arm solely with their thoughts. And the feds are pushing on the biological front as well, launching a 20-year partnership with research facilities and computer vendors to develop “ultrascale computing.” Vast in scope, the ultrascale umbrella includes everything from schemes that use neurons as computing circuitry to massively parallel computing projects, with processors sited in multiple locations. In other words, the government wants biocomputing. And what the government wants, the government gets. Some predictions are just that easy.