Scientific American has a blog post this week about the IEEE’s 125th anniversary event. During the event one of the scientists present commented on his DARPA-funded research to reverse-engineer the brain’s computational abilities
Computers are lauded for their speed and accuracy, but they don’t hold a candle to the human brain when it comes to tackling complex mathematical problems, Dharmendra Modha, director of cognitive computing at the IBM Almaden Research Center, said at today’s event. The Defense Advanced Research Projects Agency (DARPA), the U.S. Defense Department’s research arm, last year gave Modha and his colleagues $4.9 million for a project called “SyNAPSE,” through which they are trying to reverse-engineer the brain’s computational abilities to better understand its ability to sense, perceive, act, interact, and understand different stimuli.
“We have no computers today that can begin to approach the awesome power of the human mind,” Modha said. A computer comparable to the human brain, he added, would need to be able to perform more than 38 thousand trillion operations per second [[http://www.petaflop.info/]] and hold about 3,584 terabytes of memory. (IBM’s BlueGene supercomputer, one of the worlds’ most powerful, has a computational capability of 92 trillion operations per second and 8 terabytes of storage.)
And those numbers are probably “raw performance” numbers for the brain. There are those who hypothesize that even if we can bring that much raw processing power together in a computer it still won’t be able to achieve all that humans can achieve, because we have not yet fully characterized the nature of memory, let alone consciousness. We won’t have to wait long to find out: Modha predicts that we’ll be able to simulate the workings of the brain by 2018. All of this may have more than esoteric implications
In addition to boosting computer performance, enhanced understanding of the brain will enable people to communicate directly with machines, whether they are robots or mechanized prosthetic limbs. Primates have already proved that such brain-machine interfaces are possible, Miguel Nicolelis, co-director of Duke University Medical Center’s Center for Neuroengineering, said during the conference. The researcher and his colleagues last year successfully implanted electrodes in the brain of a monkey in North Carolina that enabled him to control a robot on a treadmill in Kyoto, Japan.
Well, my brain can definitely not resolve large linear systems at petaflops rates, the way even small clusters do by running HPL.
The “raw power of human brain” can probably not be applied to the same range of problems computers do. And vice-versa.
Kilian – along the lines of “a hammer makes a terrible screwdriver?”
Yes, I agree. I’m not sure what the drive is to coax a computer to act like a brain. We have 6 billion brains already. How about we build something that we don’t already have 6 billion of?
Are they talking about the conscious or unconscious power of the human brain? Eg; processing vision which is very complicated compared to processing mathematics.
With that in mind the only thing that computers are better than us at is maths. And some researchers suggest that we are able to retrieve information much faster than conventional computers. (But with some exceptions, of course)
It is hard to say anything about raw power of human brain in computer terminology. It doesn’t work like conventional computer. It’s like comparing apple with pear.
I think real power of human brain comes from optimization. It is highly optimized on what it does.
For-example, picture processing. When we look at a picture we don’t see all points of the picture in high detail. We see in high definition only at the focus point. Rest of the picture is seen in blur.
But we move our eyes quickly and see the important and critical parts of the picture and process it with minimum amount of power. Then we define what is in the picture a human, a tree or a car.
Even if we have more raw power than a super computer that’s why we can’t do mathematical calculations faster than a simple pocket calculator.
Kilian,
Your comment:
“Well, my brain can definitely not resolve large linear systems at petaflops rates, the way even small clusters do by running HPL.”
Actually your brain has that capicity. Someone created and taught that computer how to do that. It didn’t teach itself. Its brainless without a human brain.
The Brain has to keep people alive, the majority of its power goes into running organs and the visual information. That’s why we can’t accurately calculate numbers as well as a computer can. If we ever develop the technology of the human brain. It will be able to store vast amounts of information. Imagine the entirety of human knowledge in a single brain sized package. Every hard drive in the world ever fitting in 20 human brain sized drives.
the brain is indeed powerful. I think you cant really compare the power of a brain, and a computer. They are two different things and should be looked at as such. Although where I stand on this topic is that the brain processes things like images, sounds, smells and textures, while at the same time keeping the body in check. If the entire brain was dedicated to mathematics, it would be faster than any computer
Well, if my brain wasn’t occupied with constant monitoring of body temperature, sight, hearing, CO2/O2 monitoring, hormone level monitoring, respiration… THEN I would be able to compute million times better than a computer… I need to put my brain in a jar! Also, it wastes only 20 watts, super efficent!
Also, don’t forget those times when you get some idea, you suddenly get it, you KNOW how to do it, but when you try to write all the steps it takes a long while… So, what I’m saying that when a brain is familiar with something and really needs it, it restores it it 0.000000001 seconds!
the brain is more like a multi core system with a few thousand cores. It primarily differs from a PC in that most PCs have 4 cores or fewer and most algorithms are not inherently parallel. However, even the most basic neural operations are executed in parallel.
The supercomputer is fast – at what it does, what it is designed for. The human brain, though, has many advantages. It LEARNS. And does things automatically, sort of, after the eureka moment that locks in the ‘tricks’ it has learned. The short-cuts. When – someday, they learn how to make a computer mimic this way of doing things… then.. then we’ll have true, ‘supercomputers’. We could build androids.. that thought about how they would handle each task.
Think about our vision. Our eye ‘sees’ wavelengths of light, reflected off of objects. These wavelengths are not even focused by our eyes, into coherent images ! Uh-uh. Our eye is terrible at focusing light. It is a single lens, and that means we ‘see’ a blur of colors. Our brain sorts things out. That is VERY cool !
The brain is amazing for the size vs power as in supercomputer: You can fly a plane while listening to music and enjoying the taste of a candy bar. While the brain be busy running your body and you know nothing about what it is doing in your body. IMHO