Supercomputer to brain-storm the human brain
Photo by LeRoy N. Sanchez, Los Alamos National Laboratory, US DOE’s National Nuclear Security Administration.
A supercomputer with Petaflop performance fills a whole room.
Engineers have created the world’s fastest supercomputer, at Los Alamos Labs, USA, at a cost of $US120 million.
Why all this ‘crunch’ power? One purpose is to run PetaVision, a program that simulates the operation of the brain’s processing of visual information from the eye.3 For example, you might see a bird flying in the sky, which you immediately recognize as a bird. It takes such a ‘Petaflop’ supercomputer to do a similar thing.
According to the article about this,1 the human brain has 1012 synapses (nerve connections) involved in processing the information involved in seeing, so that mimicking its operation requires a supercomputer with Petaflop performance. This computer fills a whole room (see picture)—it is just as well that our brains don’t have to take up so much room!
Basic brain facts
A recent article outlined some astounding facts regarding the human brain:4
- Estimated total number of nerve cells (neurons): 100 billion.
- Number of connections in the brain: 500 trillion (5 x 1014).
- Number of new nerve connections made every second: 1 million
- Processing capacity: 100 trillion instructions per second (1014).
Neuroscientists do not yet have much idea how the brain works. The latest ideas involve it working like a massive probability calculator, making predictions and progressively reducing the errors on those predictions.3 So, for example, when listening to someone talk, we are always guessing what they are going to say next—and this can get in the way of hearing what the person actually says! This also enables us to make sense of a distorted voice over a noisy phone line. But just how the brain does those probability calculations is a mystery.
It is no mystery to our Creator, of course. When we think of how incredible our brains are, we should spare a moment to think, using those brains, just how incredible the One who designed our brains must be!
Grant B., Australia, 20 Sept., 2010
Thank you for a very interesting article. Obviously such a large computer must require gigabytes or terabytes of code to run not to mention the amount of data that was required to build it in the first place. The program which builds and runs the human ‘machine’ fits on 3 billion base pairs each coding 2 bits or about 750 MB.
Obviously the only way to fit this much information in such a tiny space is to use data compression ratios in the range of 100–1000:1. I don’t see how a programmer could build a heavily compressed program by trial and error de novo. How do you evolve a program where every element is multitasking perhaps 1000 fold?
Dominic S. Australia, 20 Sept., 2010
Contrary to Grant B’s comment, I don’t think the PetaVision software would involve “gigabytes or petabytes of code”.
I surmise it would consist of a component for simulating a neuron in the visual cortex and a component for simulating a connection between neurons, plus a component for directing the overall simulation. The first two components would be duplicated so thousands of copies of the same algorithm are working together. The amount of code that would be written to achieve this task would not be especially large (nothing like the size of a typical desktop operating system). The point of having such a ‘large’ computer is not to run a large amount of code but rather to run a repetitive process many many times.
The same applies to the brain: DNA doesn't have to encode gigabytes or terabytes of information about the processes of the visual cortex because it’s a collection of replicas of the same type of cell. Also, the PetaVision software is simulating the electrical processes of the synaptic connections, which is something that doesn’t have to be stored in DNA since that’s all handled by physical/chemical processes (i.e. that’s handled by the universe).
That’s not to diminish how amazing the human brain is, of course. Plus of course God designed all the ‘natural’ physical/chemical processes that allow all this to happen, which we are simulating/approximating. And those same things allow supercomputers to exist and operate...
Author's comment: Indeed so. The marvel is that somewhere in the DNA resides an algorithm that specifies how brain connections are to happen in response to external stimuli. There could be different algorithms for different functions (sight, hearing, touch, memory, etc.). No one yet knows. Our article on Jesus restoring the sight of a blind man in two stages touches upon this issue.
Wildee R. Philippines , 20 Sept., 2010
Thank you very much for this article. This reminds me of life saving devices like the heart and lung machine or the liver dialysis apparatus. Machines that were copied from the original to simulate and/or temporarily replace vital organs when they begin to fail. Those equipment are bulky, expensive, inefficient & very much a far cry from the original organs they are meant to replace. This is not even considering the manpower, brain power, and time it took for men and women to design, assemble, test and run them. With all our manufacturing capability these days it would be no less than 100 people involved in making the individual parts of even one equipment from scratch design to final implementation. I hope thos! e who are crying “no design” or “no designer” would stop insulting the Master Engineer, and consequently the people who work on these types of equipment, by even conjecturing that these machines just came about without any thought to them.
Preston G., USA, 21 Sept., 2010
I think Grant is closer to being correct than Dominic. For interesting results you wouldn’t simulate only thousands of neurons etc, you would simulate as many as your processing power, memory, and bandwidth allowed. And you would obviously do it all in parallel, which would require enormous amounts of memory, processors and bandwidth since billions of instances of the code must be in memory at the same time.
It has been said by some geneticists that the more they learn the more they realize they don’t know. In the field of genetics scientists have realized that some non-DNA structure affects the decoding of the DNA in such a way that diet actually can affect the offspring. I give this as an analogy to the brain and its complexity. I think that people researching brain function are going to learn that it is much much more complex than they currently assume. My guess is that it is many orders of magnitude more complex than they think.
But even with almost unlimited processing power you can’t necessarily reverse engineer a system as complex as the brain. NOAA and NWS and NHC and other govt agencies tasked with predicting the weather have a lot of supercomputing power available to them. And even though weather seems infinitely simpler than a biological system such as the brain, they can’t predict the weather accurately even a day in advance. That being the case, if the Lord’s patience continues and he does not return for another hundred years, I think articles just like this one will be just as applicable then as now.
Mike C., USA , 21 Sept., 2010
The massive probability calculator is mostly made massive, I believe, by the storage of patterns—both visual and auditory. It is fairly well understood that we can look at incomplete visual patterns and “fill in the blanks”—something that no doubt has to do with survival. These pattens are stored in various phases of our learning and a re reinforced by the circumstances that allowed their storage. Many patterns we see are repaeats, and the brain knows to ignore and dump these.
You see other “mysteries” in the animal kingdom, such as the ability of birds or whales to migrate thousands of miles—has to be pattern recognition (and in a tiny brain). Even though our brain is seemingly large and packed with data and computing power, there is still room for the Holy Spirit who guides us in our more difficult decisions. :-)
Joshua T., Australia , 1 Oct., 2010
Very good article ... Just show’s how awesome and great our Creator is...
I guess this article just proves that it will be a LONG time before anyone will be able to get an artificial brain in my head ... (joke)
Imagine trying to transport your supercomputer of a brain anywhere ...
References and notes
- World-record supercomputer mimics human sight brain mechanisms <www.sciencedaily.com/releases/2008/06/080612140031.htm>, 15 June 2008. 1 Petaflop = 1015 floating point operations per second. Return to text.
- A Dual-Core Intel® Xeon® Processor 5100 Series has a processing power of ~2 Gigaflops (2 billion floating point operations per second) <download.intel.com/products/processor/xeon/dc51kprodbrief.pdf>. Return to text.
- The vertebrate retina actually pre-processes the visual signals before sending them to the brain. See Sarfati, J., By Design, Creation Book Publishers, Brisbane, Australia, pp. 27–28, 2008. Return to text.
- Huang, G.T., Essence of thought, New Scientist 198(2658):30–33, 31 May 2008. The figures are all estimates, so they might differ from estimates made by others. Return to text.