Do you learn best by doing? It’s not far-fetched that in the not-too-distant future your computers will as well, based on cognitive computing systems that simulate human thought processes and learn to become more accurate through experience.
Taking a clue from the human brain, scientists at IBM are working on a new cognitive computer chip likened to right–brain, sensory pattern-recognizing learning models. Known as TrueNorth, this neurosynaptic computer chip is an event-driven microprocessor with supercomputer capabilities that’s roughly the size of a postage stamp and is central to IBM’s cognitive computing program.
The chip uses just seventy milliwatts of power, which is approximately the energy equivalent of a hearing-aid battery. The technology is an approach different from the traditional von Neumann architecture in use since 1946.
What are the real-world applications where a cognitive computing chip such as TrueNorth would have an advantage? Because of its small size, low power consumption, and ability to do cognitive-type processing, search and rescue operations, glasses-type devices to help people with vision impairments or blindness to navigate, self-driving cars, or diagnostic tools in medicine are just a few scenarios.
Calling TrueNorth “a direction and not a destination,” Dr. Dharmendra S. Modha, founder of IBM’s Cognitive Computing Group at IBM Research, wrote in his blog, “Let's be clear: we have not built the brain, or any brain here. We have built a computer that is inspired by the brain. The inputs to and outputs of this computer are spikes. Functionally, it transforms a spatio–temporal stream of input spikes into a spatio–temporal stream of output spikes.”
The announcement was made in an article published in Science magazine. From the abstract:
Inspired by the brain's structure, we have developed an efficient, scalable, and flexible non–von Neumann architecture that leverages contemporary silicon technology. To demonstrate, we built a 5.4-billion-transistor chip with 4096 neurosynaptic cores interconnected via an intrachip network that integrates 1 million programmable spiking neurons and 256 million configurable synapses. Chips can be tiled in two dimensions via an interchip communication interface, seamlessly scaling the architecture to a cortexlike sheet of arbitrary size. The architecture is well suited to many applications that use complex neural networks in real time, for example, multiobject detection and classification. With 400-pixel-by-240-pixel video input at 30 frames per second, the chip consumes 63 milliwatts.
In the video below, Bill Risk, one of the IBM research engineers, discusses the potential of cognitive computing.
The TrueNorth chip, being developed by IBM in collaboration with Cornell Tech and other universities, is funded by the U.S. Defense Advanced Research Projects Agency. If you are interested in collaborating with IBM Research, here’s an online information form.
