Slashdot Mirror
Effort to Create Virtual Brain Begins
bryan8m writes "An IBM supercomputer running on 22.8 teraflops of processing power will be involved in an effort to create the first computer simulation of the entire human brain. From the article: 'The hope is that the virtual brain will help shed light on some aspects of human cognition, such as perception, memory and perhaps even consciousness.' It should also help us understand brain malfunctions and 'observe the electrical code our brains use to represent the world.'"
with more details:n 2005/tc2005066_6414_tc024.htm
http://www.businessweek.com/technology/content/ju
From the article:
In other words, one day they hope to simulate a whole brain, but to begin with they'll be modelling the behaviour of a particular neural unit - with physical data derived from many, many slices of mouse brains.
In terms of deciphering the behaviour of relatively large numbers of neurons, it could be incredibly useful (and once the model is tuned would mean fewer messy, difficult and unpleasant experiments involving live animals, brain electrodes and whatnot) - but it's admittedly only a small first step toward modelling a whole brain of any species. Still, it's one of the necessary building blocks - and any moral issues are left as an exercise for the reader...
Tedious Bloggy Stuff - hooray?
All it takes to simulate a human brain is 22.8 teraflops? I thought I was smarter than that.
You are.
According to the Business Week article this thing will be simulating about 10 thousand neurons. The human brain has about 100 billion neurons. This will be simulating a small section of cortex, not an entire brain. The goal seems to be to understand how cortical columns work, not to create a simulated mind. They actually will not even have enough "neurons" to match one human cortical column, but will probably still learn alot about the circuitry....
Looking at this title and having already read a fair amount on neural physiology I thought, we do not have enough information to do this yet. Then I read the article and it is a ten year long project, and possibly for a mouse brain (clarification would be nice).
Again from the article:
Sounds like they'll use the data from this first phase to develop a simplified model of how networks of neurons behave - more of an empirical simulation rather than a from-scratch physical one.
Could be slightly cheaper in terms of computational power, but what are the philosophical implications of neurons which aren't directly based on physical simulation?
"I think I think, therefore I possibly are..."
Tedious Bloggy Stuff - hooray?
Wow, did you really just link to Mentifex's page? For those not familiar with him, he's an infamous kook from the early days of Usenet who spammed newsgroups claiming (and still claims to this day) that he's "solved AI" and implemented it in Forth and JavaScript. More recently, he's expanded onto places like slashdot.
There's a fairly extensive FAQ on him here:
http://www.nothingisreal.com/mentifex_faq.html
Not really, at a basic structural level, biologically they're almost identical, it's just a question of scale.
Schizophrenia has nothing to do with so-called "Dual/Split" personalities. Look it up
This was all covered back in the late sixties/early seventies by the great Donald Michie http://www.aiai.ed.ac.uk/~dm/dm.html If only there had been the processing power back then. The project was stopped because 'computers will never be powerful enough' such is the foresight of civil servants.
init 11 - for when you need that edge.
You're talking out yer butt on a couple of issues. First, they are not simple. Their many axions each have many dentrites. Their responses change depending upon the hormone bath that they live in. Second, they do indeed 'morph' throughout life. They can even repair. This is especially true of the dendrites.
You're pretty correct on the wiring, although not at the level you wrote. The basic connectivity and structure is known, but each and every brain is wired from experience, not just birth.
It's worth trying, and we will learn a lot regardless. We just won't learn as much about the brain as one might think.
(chopped from various web sources)
Some of the most successful early computers were analog computers, capable of performing advanced calculus problems rather quickly. Before digital computers became the mainstay of computing, analog computers were quite common. Analog computers use varying voltages and currents to represent variables, and various types of amplifiers to represent factors in differential equations, with the result being a final voltage or current that can be read out on a meter or graph. Analog computers were heavily used in process control situations, such as calculating the correct aiming of the big guns on board a battleship. Many variables had to be considered simultaneously, including the position of the ship, the position of the target, the type of ammunition, the wind and other weather conditions, the constant motion of the ship from the action of the sea, and myriad other variables. The analog computer would simultaneously combine all of these variables to generate a real-time result that would control the large servomechanisms that aimed the guns to assure that their ordinance would be delivered accurately to the target.
They were,however, a real bitch to sort out. So the computer world focused upon digital designs, which , it turned out, were a lot easier to do.
You are in a twisty maze of processor lines, all alike.
There is a lot of hype here.
Hold your horses! There is abundant evidence that single neurons can perform more complex operations than a mere 'sigmoid fuunction'. That is a working approximation that can be useful from the point of view of simulations but that is all.
Single neurons can potentially perform computations at the level of the of the passive cable equation. At the level of active membrane properties when added to those passive canle equation solutions. At the level of genetic instructions becoming activated in the nucleus and dendrites in response to activity. And finally the plasticity or learning rules that neurons use are not only computational very important but probably quite varied from brain region to region. Spike timing dependent plasticity for example allows the brain to pick out persistent correlations within highly noisy inputs. None of this is included in the impoverised neural-network viewpoint of 'sigmoids'
The real question is why are they doing this? Markram is a top researcher and knows what he is doing. But i quesiton the motivations of big blue. i wouldnt be suprised if they didnt give two hoots about the science but rather are only doing this so that they can get the kind of publicity that posts on slashdot bring. Remember 'Deep Blue'? Lets hope they dont treat Markram like they did Kasparov
for those of you who didn't get that joke.
my karma will be here long after I'm gone
They were,however, a real bitch to sort out. So the computer world focused upon digital designs, which , it turned out, were a lot easier to do.
A key factor is that analog computers are inherently lossy; components aren't precise enough to make a large analog computation as the imprecisions tend to add up...
And then there's the whole Turing concept of code as data. Analog computers were "programmed" by adding and subtracting components; software as bits is a lot more mutable. Even so, with the appropriate switching devices, an analog circuit that's programmable is theoretically possible.
But why bother when digital is so much more precise?
On the flip side, analog computers STILL see some life in minor subsystems everywhere. With proper design they happen to be quite handy for feedback-control applications...
I am disrespectful to dirt! Can you see that I am serious?!
But why bother when digital is so much more precise?
Because quantiztion and roundoff error play HELL with derivatives. Bigger, faster, cheaper digital computers had to be developed and better algorithms discovered before digital could take over the job. Once that had been done, digital's flexibility won out.
Analog computer technology was an outgrowth of audio and radio, and developed quickly during and immediately after WW II. A couple dozen components would make the fundamental building block, which could do an accurate computation (weighted sum, integration, differentiation, or something more complex) at kilohertz to megahertz rates. A similar number of components, as a digital device, could make a couple flip-flops processing a bit at about the rate the op-amp could do the entire computation. Noise and offset could be controlled, and taken into account. (In a feedback system, as in the real-world device being modeled, offset and noise are suppressed by the feedback if the system is stable.)
Analog computational technology is STILL in heavy use - at high frequencies, and at the edges of digital systems. (Digital techniques are just starting to take over some of the functions of, for instance, radios.)
The main reason digital wins out is that the computational elements are sufficiently immune to noise that they can be miniatureized and placed close together without misbehaving. When you get enough orders of magnitude cost reduction from that, you can throw enough of them at an analog problem to get an acceptable answer for less money and engineer time than you'd need to spend doing it with purpose-built or purpose-wired analog parts. Then digital wins.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
Schizophrenia was named for the apparent split between the emotional state, or affect, of the patient and the patient's surroundings. I think its a bit misleading to say the schizophrenic lives in a world in which the real and unreal are not differentiable. Its more the case that thought processes are poorly controlled, and delusional, disordered, psychotic, thinking cannot be controlled. A runaway mind seeking its own solutions.