(Page 2 of 2)
"It's very exciting," says Terrence Sejnowski, head of the Computational Neurobiology Laboratory at the Salk Institute, in La Jolla, CA. "The technology has matured to the point where it's possible to think about large-scale simulations." For example, Sejnowski studies how the thalamus, a brain area thought to relay and integrate information from different parts of the brain, interacts with the cortex. "We can currently do small simulations of hundreds to thousands of neurons, but it would be great to be able to scale that up," he says.
The million-neuron grid will have a processing speed equivalent to 300 teraflops, meaning that unlike computer-software simulations of the cortex, the hardwired silicon model will be able to run in real time. "Instead of running a thousand software instructions, it's just current running through transistors, just like real neurons," says Boahen.
Of course, the project will be a challenging one. "They will have to get a large number of chips to work together," says Douglas. "To put together a structure on the scale Kwabena has in mind--no one has done that yet." But it could become a turning point in the field. Douglas likens the current state of neuromorphic engineering to the early stages of computer-chip design. "People had been working on different types of logic gates, but it took a whole different worldview to build computer chips," he says.
Engineers ultimately hope to use the information generated by the silicon cortex in a variety of ways--to build better neural prostheses, for example. "The real-time aspect of this technology allows us in principle to interface the silicon cortex with the real cortex or brain," says Gert Cauwenberghs, a neuroengineer at the University of California, San Diego. "There is the promise, at least in the future, to build a prosthesis to replace some lost motor function or sensory function."
Researchers hope a microchip that mimics the basic functioning of the brain could perform complex calculations while using little power.
The first hybrid memristor-transistor chip could be cheaper and more energy efficient.
A new electronic device could lead to denser, faster kinds of memory, and processing chips that act more like the brain.
What sort of long term memory is being built in? Will this artificial cortex be able to rewire the neurons? Store information within neurons? These are the sorts of things that make the real cortex different from the retina.
software modelling of connections
My understanding (coming mostly from a recent New Scientist article on this topic) is that rewiring of neurons or changing the strengths of connections between neurons will be modelled with a database run basically run by a normal computer chip.
Intel recently build a processor that could process at 80 teraflops. Its high time these sort of projects started thinking big.
Take care. Raw flop counts are pretty meaningless. In this article, their term should be "TeraOp", since hardware-based NNs don't do IEEE floating point operations (FLOPs).
Also, large numbers of CPU ops/second are an easy claim when you ignore RAM bandwidth. To wit: a CPU with 64 bit registers + 3 GHz clock = 192 GigaOps. Multiply that by 3-4 functional units per CPU, and your laptop approaches a TeraOp.
So is every three year old PC a supercomputer?
Guest (Todd)
Actually it is an 80 core processor that functions at 1 teraflop.
According to Ray Kurzweil, a desktop around the year 2020 should have the computing capacity to function at the speed of the human brain (2*10^16 cps).
There also have been models of the cerebellum, not just the inner ear.
I have a model of the entire brain on my desk! It is a little rubber thingy I can squeeze when feeling stressed...
... or when feeling discomforted about having been too sarcastic in a comment to an article in a magazine.
Not entirely original or sufficient
I believe that Steve Furber at Manchester University announced some time ago a similar scale project. Like that project I anticipate that the results will be limited. Modeling the electrical properties of the brain disconnected from other physiological phenomena seems clearly insufficient.
Interesting Article and a Tiny Mistake
I believe it should read either "a better [...] prosthesis" or "better [...] prostheses". Nitpicky, I know, but I am a bit OCD about grammar.
Guest (vznuri)
what kind of algorithm is the brain
computing? maybe something hebbian, or
similar to the SVD, singular value decomposition.
check for & discuss related cutting edge developments in algorithmics & mathematics at
the theory-edge mailing list,
http://groups.yahoo.com/group/theory-edge/
If the..(people).. who are so concerned about cloning and spirituality would give it a break...
Make a real brain like the rat heart.Then the testing could be better and easier..
Manufacturing in the United States is in trouble. That's bad news not just for the country's economy but for the future of innovation.
This document is part of the “How-To Guide for Most Common Measurements” centralized resource portal. This tutorial provides a detailed guide for measurement and device considerations to take temperature measurements using thermocouples. Get an introduction to thermocouples, which are inexpensive sensing devices widely used with PC-based data acquisition systems. Also review some specific thermocouple examples and learn how thermocouples work and ways to integrate them into a data acquisition measurement system.
View full PDF >
Listen to story > 
Our list of the 50 most innovative companies, including the following:
On Twitter
Become a Fan on Facebook
Subscribe to the Feed
spicker
4 Comments
Hot air?
This is especially interesting if the nature (no pun intended) of the silicon neurons resembles that of real neurons with respect to integrative, temporal, and dynamic behavior. However, for the last 5-7 years I have been reading about projects like this that are 'underway' or where scientists are now 'thinking about starting up on the beginning of something great'. Let me know when someone has actually made some hardware that runs.
Reply
TRAyres
1 Comment
Re: Hot air?
Did you read the part that said they had replicated the cochlea (inner ear) and what, the thalmus? So now they're working on yet another part of the brain - and every year advancements are made.
What, its not good enough for you to only have modelled two parts of the brain, they have to get the whole thing done before you take notice? Let me guess: You aren't very good at investing. Just a thought.
Honestly, saying this kind of research is hot air is like saying these 'darn computers' are just passing fads.
Reply
spicker
4 Comments
Re: Hot air?
Yes, I read the part about the cochlea but didn't see the word 'replicated' only that they 'modelled' it, which I interpret as 'a model' not 'the model'. As far as I know people are still building various models of the cochlear - but I haven't heard of people claiming they replicated it.
Do not mistake, I think it is great that people model the various parts of the brain. However, one needs to carefully look at the assumptions and conditions set up for the model.
When Kwabena Boahen is planning to create 'a silicon model of the cortex' I think it is a bold statement to claim that he replicated it. No doubt he will find interesting answers and I am eager to see the results but also hesitant to expect the final solution.
When a transistor models a neuron, it models parts of its electrical properties and perhaps also wiring pattern. But seldomly, it takes into account all of its intricate chemical and physical functionality. The temporal aspects stemming from deprivation of neurotransmitter, enhancement of synapses, gene expression leading to additional receptors and the like, are most often not inherent in the silicon models presently suggested and I would question whether they are part of Boahen's but will appreciate it if it is the case.
No doubt Boahen's model will bring further understanding but I miss to see why this attempt is more special than the silicon attempts made previously.
Reply
NNemec
1 Comment
Re: Hot air?
Actually, where would you draw the line between model and replica? As I see it, the only thing that could claim to be something like a replica would be a atom-by-atom simulation. Anything that simplifies this, would already be a model.
Of course, one can try to model at various levels of refinement, but still everything that you put into the computer and call it "neuron" will be a model, unless it is a atom-by-atom simulation.
Obviously, such a complete simulation is impossible with todays and probably also next week's technology. But even more: it would not help us a bit in understanding. A replica is - by definition - exactly as complex as the original, so it is exactly as complicated to understand.
Only a reasonable model helps understanding. Scientists have to try different models to find out which aspects are important to capture the working of the original.
Reply
bshi
1 Comment
Re: Hot air?
> Of course, one can try to model at various levels
> of refinement, but still everything that you put
> into the computer and call it "neuron" will be a
> model, unless it is a atom-by-atom simulation.
Nope; the atom is still a physical model.
Reply
spicker
4 Comments
Re: Hot air?
Agree. I should not have put so much emphasis on the 'replica' word from TRAyres' response
Sorry for the confusion.
I was triggered by the title 'Building the cortex in silicon' and failed to be as brief as the reply from stevenzenith.
Reply