According to IBM, 'BlueMatter, a new algorithm created by IBM researchers in collaboration with Stanford University, exploits the Blue Gene supercomputing architecture in order to noninvasively measure and map the connections between all cortical and sub-cortical locations within the human brain using magnetic resonance diffusion weighted imaging. Mapping the wiring diagram of the brain is crucial to untangling its vast communication network and understanding how it represents and processes information.'
(Credit: IBM)Computers capable of mimicking the human brain's power and efficiency could be just 10 years off, according to a leading researcher at IBM.
According to the researcher, Dharmendra Modha, the manager of IBM's cognitive computing initiative, scientists from his company and some of the world's most prestigious universities have already managed to simulate the computing complexity of the feline cortex, a feat that could augur a day not too far off when it will be possible to ramp up to what the human brain can accomplish.
Last year, IBM and five universities were awarded a DARPA contract to work on a cognitive computing project aimed at eventually achieving that goal. Just a year later, Modha said, his team, working in conjunction with the universities' scientists, have achieved two major milestones.
The first was a real-time cortical simulation that achieved more than 1 billion spiking neurons, as well as 10 trillion individual learning synapses. According to Modha, that exceeds what a cat's cortex is capable of.
Second, the scientists created a fresh algorithm they're calling BlueMatter that is aimed at spelling out the connections between all the human brain's cortical and sub-cortical locations. That mapping is a critical step, Modha suggested, for a true understanding of how the brain communicates and processes information.
The human brain, Modha said, is fundamentally different from today's computers in power and size, and he and the many scientists he is working with are eager to learn from the brain how to build new kinds of computing architectures. Part of the reason, he added, is that as our world gets more and more complex, a "tsunami" of data is being produced and analyzing those data demands "a new kind of cognitive system, a brain-like system, to make sense of it."
To achieve the goal, Modha and his fellow scientists are combining supercomputing, neuroscience, and nanotechnology research to demonstrate what's possible. The work they've done has progressed in just a year from the granting of the DARPA contract to today's achievements.
Modha said that examples of what could be done with computers working at this scale are realistic analysis of the world's water supply systems, or financial systems. The idea is to detect causality behind phenomena, and to make those connections quickly and effortlessly, the way the human brain works. Writing such a program using today's computers would be impossible, he said, but these future computers would be able to quickly distill answers to these kinds of enormous problems.
There's no promise, of course, that Modha and his colleagues will be able to advance the difference between the power of the cat and human cortexes in the next decade. After all, there's a difference of a factor of 20 between the two. But he sounded optimistic that a decade is a realistic goal.
But regardless of the timing, the aim is clear: reverse-engineer the human brain and learn its computational algorithms. And then deploy them in a bid to solve some of the world's most complicated computing problems.
IBM and several university partners have gotten a DARPA grant to work on a cognitive computing project designed to simulate the brain's sensation, action, interaction, perception and cognition abilities. At the same time, the project's leaders will be attempting to recreate the brain's low-power consumption and size.
(Credit: IBM)IBM and five university partners have been awarded a DARPA contract to work on a cognitive computing project that will, essentially, attempt to simulate the brain's power and efficiency.
According to Dharmendra Modha, the manager of IBM's cognitive computing initiative, the idea is for him and his team to try to re-create the brain's perception, cognitive, sensation, interaction, and action abilities, while also simulating its efficient size and low-power consumption.
"The mind has an uncanny ability to integrate information from a variety of sensors, such as sight, hearing, touch, smell and can create categories of time, space and interrelationships effortlessly," said Modha. "There are no computers that can even remotely approach the capabilities of the mind. The mind arises from the wetware of the brain."
But he and his team feel that the time has finally arrived for computer scientists to at least begin to approach the brain's abilities. That's because, he said, of the convergence of three things.
First, neuroscience has made incredible strides forward. Second, supercomputing technology has reached the ability to create massive simulations in real time. And third, nanotechnology has made it possible to imagine creating simulated synapses, the very element of the brain that enables it to work at speeds and efficiencies computer scientists have so far only dreamed of.
"Together, these three trends allow us to uncover the function, computation function of the brain," he said, "while rivaling its remarkably low-power consumption and its small size."
The IBM project, just the first phase in DARPA's long-term contract, is expected to last nine months.
Already, Modha said, scientists have managed to create computing tools that have rivaled the abilities of a rat's brain.
But is it a good thing for science to try to re-create the brain's power?
"The questions are so technically challenging and it is so engaging at this point to begin to put the puzzle together, and the whole part is still not completely visible," Modha said. "I think it is premature to worry about what if."
- prev
- 1
- next
