Robot School — Will Neurosynaptic Chips Put an End to Traditional Programming?
By Bryan Bergeron
IBM’s new cognitive computing chip — which is loosely based on the structure and function of a biological nervous system — may signal the eventual end of programming as we know it. The chips — though built of silicon — are designed to learn from experience, much like the way neural networks correlate outcomes with input variables. The chip is about to be put to the test by researchers from Columbia University, Cornell University, the University of California, and the University of Wisconsin.
So, assuming IBM and the team of academicians are successful at creating practical solutions with these chips, what does it portend for robotics? For one, it means that the days of implicit calls to sensors and interrupts are a thing of the past. Perhaps programming your vacuum cleaner in five years will consist of leading your vacuum around with a choker chain — similar to that used with a dog — while you point out areas of your home that need special attention. Positive feedback might be in the form of verbal acknowledgements, and a quick yank of the choker chain could signal your dissatisfaction.
It could also mean that your automobile’s autopilot automatically and continually models your driving so as to emulate your every move in every conceivable situation. That could be a bad thing, depending on how you drive. Perhaps there will be a market for driver programs modeled after well-known race car drivers, or perhaps neurosynaptic chips trained by insurance companies. Get your 10% discount on insurance if you agree to drive the Allstate way, for example.
Does this mean you should give up on learning C++ and sit back and wait for the eager robot students to arrive? I don’t think so. The early systems — which won’t be out for a few years — will undoubtedly be hybrid systems comprised of neurosynaptic chips interfaced to traditional microcontrollers and microprocessors that follow Harvard or von Neumann architectures. Plus, these chips will be programmed the old fashioned way with C+, Basic, or some other language.
Clearly, the new chip could signal the beginning of an era of robotics ‘for the rest of us.’ Virtually no learning curve and no need for computer literacy will certainly lower the adoption barrier. The only issues remaining are cost, form factor, and legislation. The cost of a robotic home nurse might initially be affordable by only a select few, but with time, they’ll be as affordable as flat screen TVs. Form factor — meaning do you want a robotic dog, person, or cooking machine — will depend on where you are in life and your health.
Legislation, of course, will dictate the limits of the new technology. A robotic surgeon will likely be held up by the AMA and FDA for 15 or 20 years while it’s evaluated for medical efficacy. Scantily clad, humanoid robots will likely be illegal in some settings. I’m not sure I’d trust a robotic cabin crew on a cross-country flight, but I suppose that’s inevitable as well.
Assuming IBM’s chips live up to expectations, what do you see in store for robotics in the next decade? SV