Slashdot Mirror
Supercomputer Simulates Human Visual System
An anonymous reader writes "What cool things can be done with the 100,000+ cores of the first petaflop supercomputer, the Roadrunner, that were impossible to do before? Because our brain is massively parallel, with a relatively small amount of communication over long distances, and is made of unreliable, imprecise components, it's quite easy to simulate large chunks of it on supercomputers. The Roadrunner has been up only for about a week, and researchers from Los Alamos National Lab are already reporting inaugural simulations of the human visual system, aiming to produce a machine that can see and interpret as well as a human. After examining the results, the researchers 'believe they can study in real time the entire human visual cortex.' How long until we can simulate the entire brain?"
Impessive.
"So long and thanks for all the fish."
Perhaps the goal should be to make the visual system BETTER than ours?
Who the hell left colored drop lights laying all over the server room!?
-Rick
"Most people in the U.S. wouldn't know they live in a tyrannical state if it walked up and grabbed their junk." - MyFirs
And when this simulation claims to be conscious, what do we make of that?
I thought that supermodels stimulate the human visual system.
Visual object recognition systems have been a thorn in the side of robotics since the beginning. The other annoynace of battery power will likely be solved by the nanowire battery - therefore leaving 'sight' as the real final technological step for our lovely robots.
Extrapolating further, a human-quality object recognition system will yield results which we cannot currently imagine (let's avoid some big-brother robot talk for a second, however).
For example; I was looking at some old WWII photographs of troops getting on boat - thousands of faces in these very high-quality photographs. To myself, I thought,'Self. If all historical photographs could be placed in view of a recognition system, perhaps it could be found, interestingly, where certain ancestors of ours did appear.'
Throw in a dash of human-style creativity and reasoning and I'm certain some truly nifty revelations are to be found in our mountains of visual documentation currently lamenting in countless vast archives.
Read my Very Short "Stories"
It's nice to see progress is being made. It's scary how accurate Ray Kurzwiel's predictions seem to be, he said that by early 2010 we'll have simulated a human brain. (he's a technological analyst and author of "The Singularity is Near"). Todays desktops are faster then the super computers of the 90's. I can't wait till I'm able to get a laptop smarter then me in every way (queue joke about how stupid I am), that'll be a cool time to live in. Seems it's only a matter of decades away. Probably 20 years.
And we should call it Skeyenet.
Knowledge is power. Knowledge shared is power lost.
We are ignoring for the moment how the neural network simulators work, how they communicate amongst themselves, how they are partitioned, what sensor inputs they receive, how they are trained (that's a tough one), etc. This will turn out to be extraordinarily difficult unless some very clever people mimic nature in very clever ways.
Well, at least the hardware is there.
...are we scared yet?
Why not just setup another 'distributed' project where we all donate cycles and simulate the brain?
Should be enough of us out here i would think.
---- Booth was a patriot ----
From TFA it's not very clear what this simulation achieved. It was code that already existed and as far as I understand it, it was used to validate some simulation models of low-level biological vision.
However his simulation did not necessarily achieve computer vision in the usual sense, i.e: shape recognition, image segmentation, 3D vision, etc. This is the more cognitive aspect of the visual processus, which at present requires a much higher level of understanding of the vision process that we do not posess.
FYI the whole brain has already been simulated, see the work of Dr izhikevich. It took several months to simulate about 1 second of brain activity.
However this experiment did not simulate thought, just vast amounts of simulated neurons firing together. The simulated brain exhibited large-scale electrical behaviours of the type seen in EEG plots, but this is about it.
This experiment sounds very similar. I'm not all that excited yet.
First I want to say that this whole level of brain modeling is really cool. However, there are, of course, different levels of "interpretation" I don't think that this computer will be able to achieve a human level of interpretation simply by modeling the visual cortex.
Even "interpretation" at the second level above (which it seems the "roadrunner" might be able to model) require a lot more, for humans, than just the visual cortex.
In other words if we were to call into existence a floating occipital lobe connected to a couple of eyes that had never been attached to the rest of a brain we would never be able to achieve recognition/categorization let alone interpretation. If I'm wrong maybe some of you hardcore neuroscience type can help me out?
You likely suffer from mild prosopagnosia.
Machine consciousness is not something that will likely happen in our lifetime. We don't even know exactly what it is in humans, much less a machine. Neuroscience is further ahead on consciousness issues than computer science, and even they haven't turned up a great deal yet. Computer scientists and physicists haven't got a clue about this, and sometimes their drivel about consciousness and human cognition is just embarrassing to them.
There is more to science than physics!
www.iomalfunction.blogspot.com
Based on reasonable extrapolations of the rate of hardware advance, we won't be able to simulate a human brain in real time until sometime in the 2020's.
However, that is based on the previously incorrect assumption that neurons are the only kind of brain matter that is important. Now it is clear that glial cells play an important role in coordinating cognition. There are 10 times as many glial cells as there are neurons. That sets our simulation back a few years.
I think Ray Kurzwiel is way, way, too optimistic regarding the rate of progress.
It's funny that if you claim a mountain is impossible to climb they'll name it after you. But try going up that same mountain in ten minutes. Will they rename it after you? No way...
It's true that we don't know how the human brain works, yet, because we don't have all the needed tools to study it today. A caveman would never be able to understand the workings of a watch, you cannot study a watch stone tools. But each time a supercomputer beats a record we get a better tool to study the inner workings of the human brain.
After examining the results, the researchers 'believe they can study in real time the entire human visual cortex.' How long until we can simulate the entire brain?"
There are researches who believe that humans use their whole brain to "see." If that is true, the claims of these researchers are highly premature with respect to vision. Everything from stored patterns to extrapolation is used to determine what we see. Even familiarity is used in perception - that is why there is this urban myth that "foreign" people look the same. If one were to ask those foreigners, they will say all indigenous people are totally different.
I admit I didn't RTFA - but that sort of report cropping up in different places is really quite misleading in principle. While it may be true that the processing power exists to simulate networks on the scale of small parts of the brain in real time, the biological data to work on simply _does not exist_. The situation is somewhat better for the retina than for other parts of the nervous system, but seriously: Nobody knows the topology of neural networks in our brain to the level of detail required for simulations that would somehow reflect the real world situation. Think about it: A neuron is small, just several micrometers in diameter and it can form appendages of several centimeters (within the brain) in length that can connect it to several thousands of other neurons. The technology to map that kind of structure simply does not exist. It _is_ being developed, but there is nowhere near to enough data to justify calling the programs these computers run "simulations of the human brain".
How long, you ask?
Until they can emulate the quantum/holographic methods the brain employs. Keep in mind, there are some worlds-in-worlds within the physical components. Just like how metal siding can form a complete circuit around the house, the nerves of the brain form multiple networks (chemical, electrical, interference patterns, etc)
Job? I don't have time to get a job! Who will sit around and bitch about being broke and unemployed then?
I'm a PhD student in Neuroscience. Don't get too excited. This is merely just a piece of the visual cortex. How long until we can simulate the entire brain in real time? That's not likely for a long, long time, but not because we won't have the computing power (we'll have that in about 10 years), but because we won't have the entire brain mapped to simulate. In order to accurately simulate the entire brain we first have to understand each part's connections, how they work, and how they interact with the rest of the brain. Sadly, our knowledge of the brain is so primitive that I don't see us totally mapping the brain for at least another 100 years. Sound ridiculous? Ask anyone in academia in neuroscience, and they'll tell you that even tenured theories are being thrown out regularly when evidence to the contrary proves it wrong. There are even some who think we'll never fully understand the brain due to the fact that the best way to study it is in live humans and scientists are severely limited in that study by human rights laws.
In the interest of full disclosure, let me first say that I am one of the co-authors of the model that was executed on the Roadrunner, though I had nothing to do with the actual implementation that was executed (this was done by professional computer scientists, and I am a computational neuroscientist).
Let me clarify what was done, and what will be done in the future.
We simulated about 1 billion neurons communicating with each other and coupled according to theoretically derived arguments, which are broadly supported by experiments, but are a coarse approximation to them. The reason is that we are interested in principles of the neural computation, which will enable us to construct special purpose dedicated hardware for vision in the future. We are not necessarily interested in curing neurological diseases, hence we don't want to reproduce all physiological details in this simulation, but only those that, in our view, are essential to performing the visual computation. This is why we have no glia and other similar things in the model: while important in long-term changes of neuronal properties, they communicate chemically and, therefore, are too slow to help in recognition of an object in ~200 milliseconds.
The simulation was a proof of principle only. We simulated only the V1 area of the brain, and only those neurons in it that detect edges and contours in the images. But the size of V1 we simulated was much larger than in real life, so that we had only a bit smaller total number of neurons than the entire visual system in a human has. Hence we can reliably argue that we will be able to simulate the full visual cortex, almost in real time. This is what will be done in the next year or so.
When we talk about human cognitive power, we only mean the ability to look at images, segment them into objects, and recognize these objects. We are not talking about consciousness, free will, and thinking, etc. -- only visual cognition. This is also why we want to match a human, rather than to beat him: in such visual tasks, humans almost never make any errors (at least, when the images are not ambiguous), while the best computer vision programs make an error in 1 in 10 casesor so (just imagine what your life would be if you didn't see every tenth car on the road). Based mostly on theoretical arguments characterizing neuronal connectivity, and neglecting many important biological details, we may never be able to match a human (or maybe we will -- who knows? this is why it's called research). But we have good reasons to believe that these petascale simulations with biologically inspired, if not fully biological, neurons will decrease error rates by hundreds or thousands. This is also why we are content with simulating the visual system only: some theories suggest that image segmentation and object identification happens in the IT area of the visual cortex (which we plan to simulate). While the rest of the brain certainly influences its visual parts, it seems that the visual system, from the retina to IT, is sufficiently independent of the rest of the brain, so that visual cognitive tasks may be modeled by modeling the visual cortex alone.
Finally, let me add that we got some interesting scientific results from these petascale simulations and the accompanying simulations and analysis on smaller machines. But we need to verify what we found and substantially expand it before we report the results; this will have to wait till the fall, when the RR computer will be available to us again. For now, the fact that we can simulate the system the size of the visual cortex is of interest by itself.
Heh... what if they finally simulate a human brain and... he's just a normal guy. "Design a better computer for us B.O.B." "Uhhh... I don't even like computers." Or what if it turns out to be stupid? Make it 100x faster and it's just STUPID FAST. :)
Cwm, fjord-bank glyphs vext quiz
In the interest of full disclosure, let me first say that I am one of the co-authors of the model that was executed on the Roadrunner, though I had nothing to do with the actual implementation that was executed (this was done by professional computer scientists, and I am a computational neuroscientist).
Let me clarify what was done, and what will be done in the future.
We simulated about 1 billion neurons communicating with each other and coupled according to theoretically derived arguments, which are broadly supported by experiments, but are a coarse approximation to them. The reason is that we are interested in principles of the neural computation, which will enable us to construct special purpose dedicated hardware for vision in the future. We are not necessarily interested in curing neurological diseases, hence we don't want to reproduce all physiological details in this simulation, but only those that, in our view, are essential to performing the visual computation. This is why we have no glia and other similar things in the model: while important in long-term changes of neuronal properties, they communicate chemically and, therefore, are too slow to help in recognition of an object in ~200 milliseconds.
The simulation was a proof of principle only. We simulated only the V1 area of the brain, and only those neurons in it that detect edges and contours in the images. But the size of V1 we simulated was much larger than in real life, so that we had only a bit smaller total number of neurons than the entire visual system in a human has. Hence we can reliably argue that we will be able to simulate the full visual cortex, almost in real time. This is what will be done in the next year or so.
When we talk about human cognitive power, we only mean the ability to look at images, segment them into objects, and recognize these objects. We are not talking about consciousness, free will, and thinking, etc. -- only visual cognition. This is also why we want to match a human, rather than to beat him: in such visual tasks, humans almost never make any errors (at least, when the images are not ambiguous), while the best computer vision programs make an error in 1 in 10 casesor so (just imagine what your life would be if you didn't see every tenth car on the road). Based mostly on theoretical arguments characterizing neuronal connectivity, and neglecting many important biological details, we may never be able to match a human (or maybe we will -- who knows? this is why it's called research). But we have good reasons to believe that these petascale simulations with biologically inspired, if not fully biological, neurons will decrease error rates by hundreds or thousands. This is also why we are content with simulating the visual system only: some theories suggest that image segmentation and object identification happens in the IT area of the visual cortex (which we plan to simulate). While the rest of the brain certainly influences its visual parts, it seems that the visual system, from the retina to IT, is sufficiently independent of the rest of the brain, so that visual cognitive tasks may be modeled by modeling the visual cortex alone.
Finally, let me add that we got some interesting scientific results from these petascale simulations and the accompanying simulations and analysis on smaller machines. But we need to verify what we found and substantially expand it before we report the results; this will have to wait till the fall, when the RR computer will be available to us again. For now, the fact that we can simulate the system the size of the visual cortex is of interest by itself.
That's all, folks!
Between the rods and cones of the retina and the optic nerve are four layers/types of retinal processing cells. Unlike most neurons these operate entirely on inhibitory processing (rather than 85% excitatory and 15% inhibitory) and entirely on slow voltage gradient (rather than store up charge to a threshold and then fire a burst). How this accomplishes visual processing is a mystery to those of us to who understand real meatware processing. It is not likely a bunch of high powered supercomputer geeks even know this is how the visual system operates much less how to simulate it.
They way well use their XYZflops to develop a visual processing system of some sort, but it will NOT be a simulation of something that those who understand it far better than they understand it hardly at all.
If and when they get to actually trying to match the human visual system in operation (though by different processing) they'll have to figure out of to get their system to consistently guess with fairly good accuracy what it's going to be seeing 0.1 to 0.3 seconds in the future. Proof of that long suspected technique was just forthcoming in the last week or so.
There is nothing at all "intelligent" about this. It is all automated processing. Level of "intelligence" has nothing to to with visual proceses' efficacy. Anytime anyone inserts the "I" word into anything regarding computers, particulary when comparing with the human brain, they need to define their terms. Almost certainly those of us who have struggled for years with the insufficient and contradictory proposed definitions of "intelligence" in the human mind will be more than happy to fill them in on why their definitions have already been proven to be failures in humans, and why anything derived from those will not apply to system designed to provide human-looking output via entirely different means of processing.
"I may be synthetic, but I'm not stupid." -- Bishop 341-B