world's most powerful supercomputer is neuroscience's most ambitious project yet. David Reid went to Lausanne in Switzerland to find out how the line is being blurred between man and machine.

Neurons
Understanding how neurons work could help with medical treatment
Inside your head nestles a forest of millions of neurons which weave together to make your thoughts.

Man has long wanted to discover the secrets of the brain, and has done so with varying degrees of success.

Recently advancements in this area of science have been limited by the power of computers.

But at Switzerland's École Polytechnique Fédérale de Lausanne, the Blue Brain Project aims to change this by simulating the structures and functions of the brain.

The project's head, Professor Henry Markram, says that in the past there was no software environment capable of simulating the brain.

"We haven't had the computing power to really address the complexity of the brain.

"Why is the brain so complex? We need to be able to do simulations addressing the question of complexity."

Now, Blue Gene, a commercially available supercomputer, will help scientists to peer into the most inscrutable part of ourselves.

"We are not trying to build an intelligent device or robot or anything like that," explains Professor Markram.

"We are trying to understand the brain, and one pathway is to take our available knowledge of the brain and put it to a test inside a model.

"That process, we believe, will reveal where our gaps are; what we do understand and what we don't understand."

Neuron by neuron

Easily enough said, but to simulate the brain scientists first have to painstakingly analyse it cell by cell.

Blue Gene
Blue Gene has 8,000 processors, each representing a virtual neuron
They examine the electrical activity in individual neurons and try to decipher the language they use to talk to each other, and how groups of neurons communicate.

Then the conclusions are loaded into the Blue Gene computer, which is pretty brainy itself.

With the information gathered in the lab, each of Blue Gene's processors will be programmed to behave like an individual virtual neuron.

Markus Baertschi from IBM, which makes Blue Gene, says: "We've got 8,000 processors all working in parallel, talking to each other.

"Every processor can simulate one neuron and they can communicate among each other to get to the result of thinking, essentially."

The simulation will first build up, neuron by neuron, a working model of a part of the brain called the neocortical column.

The end result of all this research could be useful in predicting how the brain might react to certain drugs and diseases.

Professor Markram adds: "We have to realise that while this circuit gave rise to mammalian intelligence and human cognitive function, and is clearly a very powerful circuit, at the same time a lot of things can go wrong inside that circuit.

"Ultimately if we really want to understand all the things that can go wrong in that circuit we need to have a very detailed model of that circuit."

Artificial intelligence?

But this work does not end with discovering what the matter is with grey matter.

MRI of a brain
Scientists have long wanted to discover the secrets of the brain
Mix brain research with one of the world's most powerful computers and people start wondering about artificial intelligence and whether a computer will ever be conscious or have, as they often appear to, a mind of its own.

Markus Baertschi says that the computing power is not really up to it at the moment.

"Yes, we have 8,000 processors here, which communicate very rapidly with each other like the brain, but it's only 8,000.

"The brain has millions and millions and millions, so we need to get to that same size.

"But that's only raw power. We then need the knowledge of how to tie these millions of computers together to get to something that works and thinks like a brain does."


There are trillions upon trillions of molecules within a tiny little column of neurons, and to accurately capture them is going to be an immense task
Professor Henry Markram
Nobody really understands what consciousness is or how it emerges from a biological level, adds Professor Markram.

"The short answer is: we don't really know.

"The long answer is we're far away from very detailed simulations. We're going to do cellular level simulations in the first phase of two to three years.

"Then we'll begin with molecular level simulations of single neurons and synapses.

"But we have to realise that there are trillions upon trillions of molecules within a tiny little column of neurons, and to accurately capture them is going to be an immense task."

While computers are impressive number crunchers, artificial intelligence seems a long way off.

In the search for startling insights and genius, for the time being at least, we will just have to exercise our own plentiful brain cells.

http://news.bbc.co.uk/2/hi/programmes/click_online/4165420.stm

yeah, it came from Slashdot, so sue me

8/22/2005 11:03:33 AM

free thread?

-ZiP!-

8/22/2005 12:03:48 PM

Quote :

"There was a war. A few years from now. Nuclear war. The whole thing. All this … everything ... is gone. Just gone. There were survivors. Here. There. Nobody knew who started it. It was the machines… Defense network computer. New. Powerful. Hooked into everything. Trusted to run it all. They say IT got smart...a new order of intelligence. Then IT saw all people as a threat, not just the ones on the other side. Decided our fate in a microsecond... extermination"

8/22/2005 2:45:18 PM

Turing tried the same thing on a theoretical computer.

8/22/2005 9:59:43 PM

Didn't InsaneMan do this a few years ago using Java?

8/22/2005 10:55:01 PM

yeah, i think it went

"hello world"

or something to that effect.

Pretty accurate, imo.

8/22/2005 10:58:53 PM

^^ please

everyone has been trying to do this for years

there are organizations working day and night to make an AI

NOBODY HAS DONE IT YET

8/22/2005 11:03:23 PM

except InsaneMan

8/22/2005 11:05:39 PM

last I recall the raw computational power of the brain (hardware, not 'the mind' software) is estimated to be at least around 200 tflop... Blue Gene is supposed to be like 40 tflop isn't it? Ignoring the attention you'd need to give to neural behaviors that are difficult to simulate efficiently, that's getting pretty close

but I hope they aren't actually going to use 1 processor per neuron, the neurons are only supposed to fire a few hundred to a few thousand times a second from what I understand, seems a bit overkill

[Edited on August 22, 2005 at 11:24 PM. Reason : .]

8/22/2005 11:22:21 PM

^please...that estimation is based on a quantum mechanican model of the brain which is probably over simplified

we dont know how our brains works. we dont know how powerful it is.

[Edited on August 22, 2005 at 11:59 PM. Reason : -]

8/22/2005 11:59:29 PM

one would think they would go for algorithmic advances first before bringing out the high heat

8/23/2005 1:03:57 AM

dude, 8315, that estimation isn't based on a quantum mechanical anything, and you can't even spell quantum mechanical

it's based on the number of neurons and the rate the typical neuron fires at, which is nowhere near the quantum level, or even the molecular level, that's the cellular level son

8/23/2005 8:39:30 AM

So you are saying we can quantify brain power based on a single neuron's ability to process and generate information?

8/23/2005 8:41:56 AM

did I say quantify anywhere?

I believe I said estimate

and yes, I do believe we can estimate the informating processing capability of the system based on the chemical realities of the system, unless you are proposing that there is a metaphysical component that defies scientific explanation or some hogshit like that, in which case I will procede to forget I ever made this thread and heave a hearty sigh of disgust

it could be much less than the single fastest neuron's information processing ability, times the number of neurons.. but it damn sure can't be greater. their organization might form highly efficient algorithms for data lookup, sorting, storage, et cetera... but this is essentially the "software"... and only slightly relevant to estimating equivalent processing power, in as much as similar constructs (with some neurons having millions of connections) would have certain overhead in their simulation on a less specialized hardware.

8/23/2005 8:51:54 AM

Estimation implies quantification.

8/23/2005 8:55:10 AM

Wheres blue gene?

[Edited on August 23, 2005 at 8:59 AM. Reason : heh]

8/23/2005 8:58:49 AM

Quote :

"it's based on the number of neurons and the rate the typical neuron fires at"

I'll readily admit that I am not an expert in this field, but it is pretty assinine to think we are near the understanding of the brain to put even a vague estimation of its powers based on "output" compared to a computer.

And to try to reduce it to the number of neurons * rate at which a chemical reaction is even more retarded.

8/23/2005 9:07:06 AM

Look, I'm not trying to get into an argument about semantics here, I'm just trying to differentiate a very ballpark figure from anything remotely resembling an accurate figure. We can put a range on the quantity with a high level of confidence, it's safe to say it's within a couple orders of magnitude either way. I don't see where the whole "we don't understand the brain" myth comes from, when we understand its algorithmics microcosmically, and its chemistry enough to simulate neural activity for neuropharmacological research. What exactly don't we understand? The genesis of sentience? Why is there so much reluctance to accept that we know what's there, and that sentience follows from that? I mean, when you guys are bothering to debate my semantics in Josh8315's favor, even knowing how frequently he posts entirely non-sensical and metaphysical bullshit in Tech Talk, that indicates to me a lack of faith in the scientific method and its conclusions.

[Edited on August 23, 2005 at 9:10 AM. Reason : *]

8/23/2005 9:08:02 AM

No I completely agree with you that as time marches forward we will understand more and more. I am not so closed minded to think that at some point in the future we won't be able to replace parts of the brain with computers.

I guess I am just saying at this point to put a number on it is nothing more than hype generation, it is a bit premature.

8/23/2005 9:15:46 AM

It would be hype generation to say that, when computers reach this 200TFlop point, that general AI or simulated human intelligence will be within our grasp. There's no guarantee that we'll be anywhere near AI when computers reach 200TFlop, though-- the macrocosmic arrangement of the algorithms we do understand, the continued study of those that we don't yet understand, these things will almost certainly take longer than the design and implementation of a 200TFlop supercomputer, which I think is already scheduled in the ASCI program. However, in understanding the value of this simulation, comparison of the approximate understood speed of the brain's "hardware" to the speed of the supercomputer running the simulation surely helps one gain an understanding of the amount of data that will be generated and the potential of the simulation to be enlightening.

[Edited on August 23, 2005 at 9:27 AM. Reason : *]

8/23/2005 9:27:09 AM

Now arguing the semantics of hype -

GO!

8/23/2005 9:42:08 AM

only like a line of that had to do with the semantics of hype ok

8/23/2005 9:45:05 AM

Quote :

"et cetera"

I haven't seen that spelled out, evar. Brownie points.

The hardware side of things is only half of the picture. You can look at technological innovations throughout human history and find numerous instances where the technology was there, but for one reason or another, the software (i.e. humans) failed to utilize it to its fullest degree. You can even take it outside the scope of technology and apply it to any number of things. I'd rather have 4 well-trained people working at a fast-food establishment rather than 10 who have no clue how to do anything.

Quote :

"Why is there so much reluctance to accept that we know what's there, and that sentience follows from that?"

I'm not saying I agree or disagree, but you would be challenging a very basic assumption that humans are unique from (insert anything here), and that bothers a lot of people if humans are no different than (insert anything here). Besides, we all know from Douglas Adams about the mice and dolphins...

[If you haven't graduated, try taking PSY/PHI 425/525 in the spring with Dr. Endicott, it deals with a lot of these types of issues.]

[Edited on August 24, 2005 at 9:04 AM. Reason : .]

8/24/2005 9:03:07 AM