Slashdot Mirror
Mouse Brain Simulated Via Computer
Mordok-DestroyerOfWo writes "Researchers from the IBM Almaden research lab and the University of Nevada have created a simulation of half a mouse brain on the BlueGene L supercomputer. 'Half a real mouse brain is thought to have about eight million neurons each one of which can have up to 8,000 synapses, or connections, with other nerve fibres. Modelling such a system, the trio wrote, puts "tremendous constraints on computation, communication and memory capacity of any computing platform."' Although there's more to creating a mind than setting up the infrastructure, does this mean that we may see a system for human mental storage within our lifetimes?"
Researchers ran in terror of a big cat. News at 11.
We don't use it any more, 'cause the computer keeps running away and hiding under the desk.
Assuming that the virtual mouse brain runs on Linux, I propose that we start work now on a virtual mouse trap.... The only question whether we need to develop a virtual spring, or virtual glue.
Three Squirrels
NARF $
/^([Ss]ame [Bb]at (time, |channel.)){2}$/
I opened my mouse and there was just a single chip in there. Why use BlueGene to simulate half of that?
No one is forcing you to read the textbooks that explain how your brain work. In any case, a bound on complexity was already achieved when we figured out we were made out of atoms, and how many of them.
With the continual, exponential increases in computing power that we are getting, in about 25-30 years we should have the capacity to simulate human brains. And yes, this does have a lot of consequences for how a lot of people view themselves... but already we know that we don't have free will (we make decisions before we are aware of them, for example), and we already have lots of support for reductionist viewpoints. Simulations are just an extension of that.
If you want more solid arguments for this, read The Singularity is Near, by Ray Kurzweil. He makes a convincing argument.
You can lead a horse to water, but you can't make it dissolve.
and
How can it be half a mouse brain if it has 1/1000 the number of a real half mouse brain? Their simulated neurons also had less synapses than the real thing. So is the 8000 a typo, or am I missing something?
Oh no... it's the future.
You sir, have hit the nail mostly on the head. Lately we humans are discovering something new about ourselves almost daily. The genetic link to why some of us have body clocks that are slower than others is one, genetic links to everything from sexuality to diseases. We are learning slowly that we really aren't that complex. We just didn't know that yet.
The short answer to the original question is no. The reason is that the methods used to implement the models is incapable of truly mimicking the human brain. One piece of evidence for that is the fact that we understand how computers work but not the brain. From what it appears, there is the equivalent of many computers inside our heads, each doing their own thing and communicating with the others as needed, but in very complex chemical ways as well as electrochemical. If you thought modeling planetary weather was difficult, this is orders of magnitude more difficult.
The good news is that we are trying, and from that will come many good things though I worry about what kind of damage we will do if we can figure out gene therapies that can cure cancer as well as sexual orientation. This stuff really is SciFi writers playground. We should all worry too. GM food is in your future if not already in your stomach. Perhaps next will be a special bed that you go to sleep on and wake up a good citizen in the morning?
Support NYCountryLawyer RIAA vs People
Unlikely, given that we are really no where close to even understanding completely everything about our complex brains.
Do we even want to, wouldn't that take away some of the mystery behind humans. Afterall if we can figure ourselves out then doesn't that mean that we aren't really all that complex?
wouldn't that also give us perfect explanations of people's actions making situations predictable violating free will?
afterall if society is ultimately chaotic in terms of our understanding, then wouldn't this be the ultimate control?
Don't be afraid to know more. It's coming if you want it or not. It doesn't mean a thing about free will: did you ever believe that your free will belong to your "ghost" or something? You are the sum of your parts and the interaction between them. Nothing scary about this.
As for the "mental storage" - simulating a brain doesn't mean much about mental storage. Knowing and simulating an Intel chip in a program doesn't mean you can crack open an already produced Intel chip unit and hack few more cores in it.
Plus, we already make very good use of tools to expand our mental storage: starting with notes, diaries, databases, computer knowledge systems, customer relationship programs, photos albums etc. etc.
All these act as peripheral devices to our brain, and we should expect tighter integration between the brain and those (for example a wire projecting video directly in your cortex), but nothing that "expands" the brain structure at such a low level as is hinted in the summary.
What did the author mean by that? If they are not simulating any of the actual neural structures in the mouse brain, does it mean they are just simulating a more or less random neural network with eight million neurons? I have seen reports of simulations of actual brain structures in more primitive animals years ago.
Until they can, as they say, "add structures seen in real mouse brains" there's nothing to see here, move along...
There is no connection between the simulation and human mental storage. None at all. Why the nonsensical statement in the article!
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
Afterall if we can figure ourselves out then doesn't that mean that we aren't really all that complex?
You think that making something that can figure itself out is simple?
Send email from the afterlife! Write your e-will at Dead Man's Switch.
If they can simulate half a mouse's brain, then they can surely simulate a politicians. Now we can start rounding up those scum and replacing them with computers ...
I wank in the shower.
We have a fairly good understanding of the way a rainbow is made, but I can still appreciate it's beauty. Same goes for a wide variety of phenomena.
We understand the physiological make-up of boobs, but they're still pretty interesting and appreciated by a large % of the population. Just because we understand something, doesn't make them less wonderful and amazing. Besides, most people in the near future wont bother/be able to learn about the exact way a mouse brain works, let alone a human one. So those people can still have that ignorant bliss you promote.
While it's a bit of a tangent, regarding your free will comment... Psychology does allow us to make probabilistic predictions about how populations of people will behave in a given situation. That seems to rob us of free will? But at the same time, some sort of regular predictable nature has to exist in order for us to make choices. If I can't use some sort of rudimentary psychology to predict how a girlfriend will react to my gift of a pair of tickets to the superbowl, versus tickets to the theater, then how can I be said to be choosing anything? I need to be able to predict how people will behave, or else I can't make informed choices with my own "free will"
Did it think "Cheese!" .. or was that the other half of the brain?
- Martin
For more information on the Blue Brain Project which appears to be the same, or atleast a strikingly similar project but from switzerland, click...err, that link I just placed! Here also is a good article to learn more about blue brain. It seems much more detailed than the BBC's snippit.
Groups of neurons started becoming attuned to one another until they were firing in rhythm. "It happened entirely on its own," says Markram. "Spontaneously." Insights like these are absolutly amazing. It's all such facinating research, but I can help feel a twinge of sorrow for the poor thing. the main purpose of the artificial brain, say its creators, is to make new types of experiments possible. For example, what happens when damage is inflicted on certain types of cells whose function still isn't determined? How many cells can be switched off until the behavior of the surviving cells around them becomes erratic, or the entire circuit breaks down? The poor thing is just circuits and reactions, I know, but I feel sorry that it's literally being torn apart and rebuilt all the time. It's odd, I don't feel this way in similar experiments with real mice; I guess I have a soft spot for computers...
Demented But Determined.
It's cool that we can create the basic scale of the infrastructure of a (half) mouse brain - but if we're really going to simulate a brain, we need the ability to read the contents of a real one in order to verify our simulation. Otherwise, we have little basis for saying that input X gives the sensation of movement, and would have effect/output Y in terms of changed state/response.
I wonder what the current state of neuron state reading is - would we ever theoretically be able to read the state of a brain beyond the external outputs? Could we ever get a sinlgle state that would be the 'ROM' of a person's memories and mental state, that you could place in a simulation and have that person's memories 'wake up' in a simulation? I wonder how close we could get.
Ryan Fenton
Developing simulations involves using abstractions and simplifications to deal with the fact that we can't handle the computational complexity of quantum-level simulation of an entire mouse brain.
I've seen far too many papers where people make a "simulator" for a system, without demonstrating that the simulator has any real connection to reality, and then make grandiose claims about the real system that they're simulating, based on simulation results.
Call me a cranky old computer scientist, but someone simulating a brain isn't particularly noteworthy. Showing that the simulator is accurate enough to shed light on the ways that brains work, or that the simulated mouse brain can achieve things that we have difficulty achieving with traditional computer software, and I'll be excited.
No, he was referring to Goedel's theorem whereby any sufficiently complex system is unable to describe itself. Thus, being able to understand/describe ourselves completely would mean that we are not very complex. I hold the opposite view, i.e. we will not be able to describe ourselves fully precisely because we are too complex, but Goedel's theroem might be proven wrong in the future. That'd be great news for transhumanists.
Global warming is a cube.
We used to think that the liver was responsible for anger, and the heart was responsible for love, because those are the things that seemed to react when we felt those emotions. But boy did those bafflingly complex notions fly out of the door when we discovered emotion is due to having a mass of billions of interconnected
I could go on and on and I have a very simplified laymans view of how the whole thing works.. I don't know how you can say we're starting to realize how simple we are, we're realizing how complex we are.
GM foods, by the way, haven't had their actual genomes modified, they have new genes added that create new proteins that can do things like attack insects. It's nothing as complicated as actually changing an existing gene in a useful way, which would be much more difficult because of the ways genes interact in so many ways.
// MD_Update(&m,buf,j);
However, there is an alternate theory of consciousness, based on quantum physics. It is inherently non-deterministic and cannot be modeled in a computer.
Hence, IBM's big assumption may be wrong. However, at least, the IBM experiment will tell us whether the operation of the brain is strictly Newtonian. If this artifical brain behaves differently from a mouse brain, then we would know that non-Newtonian physics is crucial to the operation of a flesh-and-blood brain.
I would imagine a Beowulf Cluster of these, but I want to be able to sleep tonight...
It must have been something you assimilated. . . .
Do we even want to, wouldn't that take away some of the mystery behind humans. Afterall if we can figure ourselves out then doesn't that mean that we aren't really all that complex?
Would it bother you to wake up one day and realize you don't have free will?
Or perhaps the soul is nothing more than chemical reactions that only came about through random chance?
Truth be told, the brain exists in a semi-logical universe where rules are applied and must adhere to the laws of physics.
The question of having free will or a soul makes no difference to how the human mind works on a chemical level. It would work regardless of how we thought on the matter (maybe just different regions) but it would still function.
So if we find tomorrow exactly how the human brain functions on an atomic level or forget the whole matter entirely, it will change nothing of how it is made and how it actually works.
And we might as well try to figure it out, because leaving well enough alone would have left us in caves thinking that fire was a bad idea.
"I am the king of the Romans, and am superior to rules of grammar!"
-Sigismund, Holy Roman Emperor (1368-1437)
However, there is an alternate theory of consciousness, based on quantum physics [quantumconsciousness.org]. It is inherently non-deterministic and cannot be modeled in a computer.
I think the biggest argument against this is that synapses do not work on the atomic level. They are made of atoms, but quantum states do not seem to overtly affect organic matter at cellular level.
Of course I could be wrong about this, but since decisions are usually the next best move it could simply be a matter of weighting what the "intelligence" applies to his rules as next best move.
The problem with General Artificial Intelligence is that "the next best move" is often open ended and too many possible choices often give our current computation a run for its money unless its put into some form of predefined rules.
The reason humans do so well is because we have certain criteria encouraging us to do things (hunger, pain, altruism, fear, etc etc)
Hence, our general intelligence goals aren't that complex (usually... to feel good about oneself and one's life) and that our true intelligence is being able to recognize things that improve upon that given a set amount of rules we know.
Which makes us very deterministic.
Even rebelling against the crowd can often be very predictable in humans.
"I am the king of the Romans, and am superior to rules of grammar!"
-Sigismund, Holy Roman Emperor (1368-1437)
- Simulated only half a mouse brain
- Ran at about 1/10 the speed of a real mouse brain
- Only ran for 10 seconds
- Only simulated generic tissue (didn't contain brain structures found in real mice)
From the article: Imposing such structures and getting the simulation to do useful work might be a much more difficult task than simply setting up the plumbing.For future tests the team aims to speed up the simulation, make it more neurobiologically faithful, add structures seen in real mouse brains and make the responses of neurons and synapses more detailed.
It's not that this isn't noteworthy, it's that mammalian brains are incredibly complex. I would be curious to see if they could faithfully reproduce a fish or reptile brain at this point.
on't be afraid to know more. It's coming if you want it or not.
John Conner: But I thought we prevented Judgement Day?!
Terminator: Judgement Day is inevitable.
"When life gives you lemons, don't make lemonade. Make life take the lemons back!" -- Cave Johnson
So was I the only one who read "system for mental storage" as meaning the transference of a human conciousness into a computer?
That's just as unlikely. People used to computer technology know that the hardware structure and the software state are two completely different things. This is why you can build a model of the hardware, feed it the state, and bang, you have a Gameboy emulator (or whatever).
But with biology, those two are intermixed. Brain saves information by changing the connections and structure itself. This means that you can build a model of a generic human brain, run it, and you have full blown AI.
But you can't feed it the state of any human being. As every human being has different "wiring", hence won't "play" in your model.
Someone mentioned Smalltalk. Smalltalk kinda works like a brain in that regard. State is structure is state.
This is completely wrong. Gödel's theorem does not state that "any sufficiently complex system is unable to describe itself." Very roughly, it (specifically the first incompleteness theorem) states that any consistent mathematical system that is able to describe itself is necessarily incomplete. And, there is no chance that "Goedel's theorem might be proven wrong in the future." It is a theorem, a mathematical truth. Not a "theory", if that's what you are confusing it with. For more info see Gödel's incompleteness theorems.
Ignoring the obvious question of whether Penrose is correct...
What makes you think this machine is not affected by cosmic rays?
Pfft...I can simulate half a mouse brain:
while(1){}
ZuluPad, the wiki notepad on crack
To
This is a very difficult, unintuitive concept, and it completely abolishes the idea that you can predict human behavior, even though you may be able to reach better and better approximations as you reach larger scales.
How does it remove the possibility of predicting human behavior? Many macroscopic processes (e.g., motions of the celestial bodies) can be predicted very well, despite quantum uncertainty. You would have to argue that human behavior is determined at the quantum level, as Penrose does, not very convincingly, in my view.
You may also consider the fact that uncertainty does not just arise at the quantum level. for example, it is very difficult to predict weather, despite the fact that quantum effects probably have little role in it. It has to do with the fact, that certain systems are very sensitive to the initial conditions and our ability to measure is limited.
Penrose is an excellent mathematician, but he's a crackpot when it comes to biology and the brain.
As for brain simulations, they almost always use randomness in the form of pseudo-random number generators. Physical random number generators are actually available and could be used, but nobody bothers because there is no conceivable way in which that could make a difference.
You made a number of spurious statements to support your thesis that IBM made a big assumption:
It is possible that brain activity occurs via the microtubules, but this has not been well shown.
Quantum physics is not *efficient* to simulate on modern computers, as the non-deterministic aspects tend to drive the model exponential. This does not prevent extremely large deterministic computers from modelling inefficiently, nor does it prevent prevent Quantum Computers from modelling more quickly (kudos to other reply who posted this point faster).
That theory of consciousness is not a particularly scientific theory. I say this since the fundamental thesis appears to be that there is 'something about consciousness' that prevents it from being possible to be simulated on a computer, as opposed to a more specific thesis. Care seems to have been taken to avoid testable claims, like the ability to solve particular classes of problem on a computer, or the ablility for a computer to pass some sort of Turing test. The heavy reliance on a slippery definition of 'consciousness' is critical. Lastly the main authors are not supporting their cases by publishing papers in decent journals, but instead by selling books and videos.
Even Penrose (of string theory fame), attempting the Lucasian argument in An Emperor's New Mind, resorted to choosing 'a mathematicians' ability to, in principle, prove any true theorem.' as the most viable testable aspect of consciousness. Since a computer will always (because of Godel's incompleteness) have statements that it cannot prove, Penrose argued that a mathematician must thus be more than any computer could be. The supposition that a mathematician's ability to freely choose between formal systems gives it the ability to prove anything is a bit of an eye popper for me, even with Penrose's 'in principle' tacked on.
Penrose followed the rebuttal well: In the same way that any computer is existing within a formal system, and thus is unable to prove certain theorems (Godel's), humans exist in physical reality - which is simulable (yes, including quantum physics) by either a large or quantum computer, given all the physics known to science at this time. This means that anything subject to known physical laws can prove no more than whatever some astronomical ultra-computer capable of simulating that subject, could prove.
The result of this painful train of thought, for Penrose, was the supposition that there must be some fundamental new physics, operating within the brain, that enables us to have the potential to solve any mathematical problem for which a solution is out there. Penrose hopes that this physics is lingering near the microtubules, but he is totally clear that it is not normal Quantum Physics, since that doesn't escape computer simulable activity.
I am not bringing up Penrose as a Straw-man - I feel he did the best job of analyzing and supporting the position. In particular, I am not saying 'blah - just Penrose et. al. not understanding what the scientific method is' In fact Penrose is well aware of the scientific method, and classifies definitions of science along a strong-weak path with strong-definitions requiring that theories be thoroughly disproven to be considered scientific. He considers his own views - namely that a scientific theory should be disprovable 'in principle', to be a weaker than normal definition.
Penrose's knowledge of computation and physics, and the quality of his arguments, far surpass the other writers in this area. He is the only fair target. Besides, his webpages have never used the blink tag. Penrose is cited on the parent's link, but it is hard to criticise the position of that linked author without having bought their video or book.
That said, I obviously disagree with his position, and that of the parent. In particular, obviously, none of this series of experiments at IBM can or will shed
Well, talk about big assumptions... I did two semesters in quantum physics as part of my electronics engineering degree. There I learned a bit about this "quantum" stuff that so many people throw around so easily.
The first thing that must be understood is that quantum effects appear in *very* small dimensions only. Quantum computing experiments must be performed under extreme conditions, a tiny fraction of a degree above absolute zero, just to get a quantum entanglement of a few bits for a perceptible amount of time. There's no way one could obtain quantum effects beyond normal chemical reactions in a human cell.
Roger Penrose, who started this "quantum consciousness" theory is a mathematician, not a physicist. He did it probably as a response to the evolving research on neural networks, such as the one mentioned in this article, based on a philosophycal uneasiness about the idea of us having a deterministic brain. He has been debunked by quantum physicists many times since he published his book.
Yet, he needs not worry. We can have a brain that's fully deterministic at a microscopic level without doing away with free will, if we assume that our brains operate in non-linear conditions.
Besides, it's not as if we had to reproduce exactly the working of living beings to emulate them. Airplanes are able to fly higher and faster than any bird without flapping their wings. At this time, we are like aircraft engineers were in the 1890s. Perhaps we will be able to find better mechanisms than used in natural brains for processing thoughts.
Hmm, couldn't you just give the simulator a source of entropy, such as a hardware random number generator? Or perhaps implement the simulator in an FPGA, and then overclock it to the point where it's just a little finicky?
Given the difficulty of distinguishing between pseudo-random and truly random numbers, I don't think that would even be necessary. I would be very surprised if we made a brain simulator with a real entropy source, which was creative, and then replaced that with a pseudo-random number generator, and the creativity evaporated.
Slashdotters are disappointing me on this one. I practically fell out of my chair when I read that half a mouse cortex has been simulated. This is an INCREDIBLE advancement. Furthermore, the fact that synchronous firing (which is probably THE most important feature of neuronal activations in the cortex) was observed is absolutely remarkable. First of all, the human brain is not a quantum computer. The scales as WAY to large, and this has been known for many years . Only crackpot websites have suggested otherwise. There is absolutely NO serious research on quantum computation in the brain. That theory was discarded several years back by serious researchers. Secondly, there are consistent theory of the mind. In particular, Jeff Hawkins Memory-Prediction framework. It is becoming increasingly evident that the mind is simply an extremely powerful predictor, which evolved to protect organisms. And third, recent research has linked synchonous firing to CONSCIOUS attention. Which means that this simulated half a rat brain is most likely the first conscious simulation EVER. This is HUGE. I cannot believe, on a forum such as this, that people would be making jokes that imply that absolutely nothing has been accomplished. This is practically the beginning of the next stage of human evolution. I could say more, but I'll let slashdotters redeem themsevles.
Setting up computational infrastructure is an important but extremely unexciting step in neural simulation. The very fundamentals of *what* we should be simulating are still largely unknown. If anyone has read more on the research please correct me, but my guess is they are running simulating 8 million undifferentiated neurons disconnected from real-world input. This isn't a "mouse brain", they just call it such because the number of "neurons" is similar.
Why do so many people refuse to entertain the possibility that they might be deterministic?
Well, it's not like they have a choice.
How to solve most of our problems: 1.Lots of nuclear plants. 2.Cure aging.
A massively parallel cortical simulator with (a) phenomenological spiking neuron models; (b) spike-timing dependent plasticity; and (c) axonal delays.
(see the actual research description here: http://www.modha.org/papers/rj10404.pdf)
Secondly, it is not necessary for a cortex to have left-right brain functionality in order for it to function. This has been demonstrated in live humans.
And third, the speed, relative to real-time, is irrelevant. It is comparatively a minor task to increase the speed of the simulation by increasing parallelization.
Now, to respond to your somewhat antiquated understanding of the current state of AI:
In addition, everything I have seen in tech press on AI since the rules based AI reasoning failures of the 80's has been neural net simulations looking for patterns, such as the mentioned synchronized firings.
Sounds like you're a couple of years behind (as would be expected on slashdot, which primarily focuses on IT and science, and not neuroscience). Let me bring you up to date a little. Spiking neural networks began to grow in popularity in the mid to late 90's. They are much more biologically realistic then most of the models used in the 80's and early 90's. Also, a lot of research has been done which points to the significance of chaotic attractors, which arising from phase-locked loops in the neuronal structure. The fact that synchronous firing is observed tends to imply similar dynamics are occurring.
Furthermore, you make the assumption that biological brains are somehow superior to simulated brains, just because they are more chemically complex. That assumption has absolutely no research to back it up. For all we know at this point all of that chemical complexity may be superfluous for evolutionary benefits (and this is direction which evidence suggests).
Aren't the neural net rules just tweaked until they get interesting behavior like that?
That's the way it used to be done, so I can understand your confusion here. I think the problem lies in the fact that people are very interested in neuroscience these days. But a remarkable amount of progress has been made. Phenomenological spiking neural networks are quite a bit biologically accurate than the "neural nets" of the 80's and early 90's.
Don't tell me you think they actually have any idea how they would simulate brain functionality.
The cortex is arranged into mircocolumns of neurons, which have a very definite structure repeating structure over the surface of the cortex. Jeff Hawkins has recently presented a very convincing argument for structure of the mind, in relation to the structure of the cortex.
Training neural nets is just something easy to do. Beats actually writing complex code, doesn't it?
If you're implying that the simulation was not complex, consider that each neuron had its own dedicated computer. And, once again, this is much more complicated than a simple neural network.
I've never seen any explanation for how either short term or long term memory works, much less reasoning or any other functionality. And that at least is something that would seem able to be modeled and explained. How does man know anything about something they have never encountered before, for example, to acquire language as a child?
Explanations for both short and long term memory have been out there for quite some time. But neuroscience is not a popular topic of discussion, partly because it can get quite complex. People would much rather be talking about the step in the evolution of Intel processors, or life