Computer science is getting stunningly close to granting the wish of the Scarecrow, not to mention the needs of the modern soldier.

The Pentagon has long sought what the Wizard of Oz could not manufacture: a brain, or at least an electronic cognition machine that operates as closely as possible to the speed and efficiency of the human cortex.

A coalition of IBM’s research institutes and severaluniversities and government labs delivered a preliminary answer Thursday to that request: a 5.4-billion transistor chip with 1 million programmable neurons and 256 million synapses. The TrueNorth chip is the size of a postage stamp and is more than 1,000 times as energy efficient as a conventional chip, according to a study published online Thursday in the journal Science.

Don’t expect to see the tiny supercomputer on your smartphone anytime soon, although the lead researcher said his team is gaining momentum in that direction. He envisions a world populated with sensors that can process data at brain-like speeds, serving such roles as guides for the blind or instant detectors of industrial toxins.

“The impossible has become possible, and the next step is to make possible real, in terms of commercial applications,” said Dharmendra Modha, head of the cognitive computing group at IBM Almaden Research Center.

Science, 8 August 2014:  Vol. 345 no. 6197 pp. 668-673 

DOI: 10.1126/science.1254642

A million spiking-neuron integrated circuit with a scalable communication network and interface

Modeling computer chips on real brains

Computers are nowhere near as versatile as our own brains. Merolla et al. applied our present knowledge of the structure and function of the brain to design a new computer chip that uses the same wiring rules and architecture. The flexible, scalable chip operated efficiently in real time, while using very little power.

Science, this issue p. 668

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.

Read the Full Text

Authors:

  • 1IBM Research–Almaden, 650 Harry Road, San Jose, CA 95120, USA.
  • 2IBM Research–Austin, 11501 Burnet Road, Austin, TX 78758, USA.
  • 3Cornell University, 358 Upson Hall, Ithaca, NY 14853 USA.
  • 4IBM Engineering and Technology Services, San Jose Design Center, 650 Harry Road, San Jose, CA 95120, USA.
  • 5IBM Research–Tokyo, Nippon Building Fund Toyosu Canal Front Building, 5-6-52 Toyosu, Koto-ku Tokyo 135-8511, Japan.
  • 6IBM T. J. Watson Research Center, 101 Kitchawan Road, Yorktown Heights, NY 10598, USA.
  • 7Cornell Tech, 111 Eighth Avenue No. 302, New York, NY 10011, USA.
  • ↵†Corresponding author. E-mail: dmodha{at}us.ibm.com
  • ↵* These authors contributed equally to this work.