An increasingly instrumented, interconnected planet demands an ultra-smart computing platform that can deal with the data produced by a plethora of sensors – much the same way as the human brain integrates sight, hearing, touch, taste and smell to process information. A recent US$4.9 million grant by the Defense Advanced Research Projects Agency (DARPA) may help IBM Research and its university partners make this multilateral processing behavior a reality for global computer networks.
Coping with a data tsunami
The digital data explosion shows no signs of slowing down. The amount of digital data is growing at a mind-boggling 60 percent each year, according to IT analyst firm IDC, giving businesses access to new streams of information. Yet without the ability to monitor, analyze and react to information in real-time, the majority of this business value may go untapped.
Imagine, for example, the business of monitoring the world’s water supply. Billions of sensors and actuators are at work in our oceans, recording parameters such as temperature, pressure, conductivity, wave height, acoustics, ocean tide and turbidity. All of these metrics are reported in real- or near-real time, creating a stream of continually changing raw data that flows at a dizzying rate. Making sense of all that input would be a Herculean task for one person…or even a hundred. A cognitive computer, acting as a “global brain,” could put together the disparate pieces of this complex puzzle and come to a logical response -- quickly and accurately.
Modeling the mammalian brain
Under the auspices of an initiative called "Cognitive Computing via Synaptronics and Supercomputing," a computing team at the Almaden Lab aims to create a universal computing platform based on the high-level computational principles of the brain. Ultimately, the team hopes to rival the brain’s low power consumption and small size by using nanoscale devices for synapses and neurons. (See text, below horizontal line, for the initiative's four key points.)
Recently, the cognitive computing team, led by IBM researcher Dharmendra Modha (pictured), demonstrated the near-real-time simulation of a small mammalian brain using cognitive computing algorithms with the supercomputing power of IBM’s BlueGene. With this simulation capability, researchers are experimenting with various mathematical hypotheses of brain function and structure as they work toward discovering the brain’s core computational microcircuits.
DARPA funding will help the cognitive computing group work on understanding what may be called the "fundamental algorithm behind the brain" -- including perception, action and cognition - to create a truly intelligent business machine.
Cognitive Computing via Synaptronics and Supercomputing: The cognitive computing initiative at IBM Research
Associated contributors: IBM Research, Stanford University, University of Wisconsin-Madison, Cornell, Columbia and UC Merced.
The initiative's 4 points:
Synaptronics: Build nanometer-scale, low power devices that exhibit synapse-like function.
Stuart Parkin (leader), Spike Narayan, Bulent Kurdi, Campbell Scott, Chung Lam (Yorktown), Philip Wong (Stanford)
Circuits: Integrate the synapses into a cross-bar array built BEOL; create neurons in CMOS circuit-built FEOL
Chung Lam (Yorktown) and Philip Wong (Stanford), Kwabena Boahen (Stanford), Rajit Manohar (Cornell)
Supercomputing simulations: Simulate the function of the electronic synapses and circuitry via BlueGene. Use data from neuroscience to create wiring diagram of the brain to guide simulations and circuit design.
Dharmendra Modha (leader), Gulio Tononi ( Wisconsin ), Stefano Fusi ( Columbia ), Brian Wandell (Stanford)
Environments: Demonstrate intelligent behavior in virtual environments.
Paul Maglio (leader), Chris Kello ( UC Merced )
Last updated on November 26, 2008