I am constantly amazed by the fact that despite how much we know about the brain, we still have not really begun to scratch the surface. I hope these two projects give us new insights, but it is important to realize it will probably be decades after the projects are done before we see any real benefits from the research.
Neuroscience is interesting because it's on the cusp of a shift. For a long time, we knew very little about the brain. Once study became feasible, we did what we could. But the field is still in its childhood. There is lots of exploratory research to do still - just, "holy shit what does this do?" Glial cells have not been studied nearly enough. I saw a talk last month about... astrocytes or microglia (oh god I'm the worst) doing some stuff we had never even thought of. So there are still total mysteries to be probed and we don't know what questions to ask.
However, similar to physics in the beginning of the twentieth century, we're hitting a point where we're going to start seeing more modeling. In a mature field, most of what is done is trying to find better models for most phenomena, as opposed to just poking things because we don't know what happens when we poke them. Neuroscience is starting to see that - more and more models to reflect the underlying world, even while new discoveries about things we never knew we never knew are made.
So the field is in an interesting position. It's straddling the line between a totally new and mature discipline. Essentially, it's in its adolescence, where it's just starting to become very mature and useful while still having that childish wonder about it.
Duke neuroscientist Miguel Nicolelis made it clear at this year’s American Association for the Advancement of Science meeting: he is not a Singularitarian. Addressing fellow scientists, he dismissed the singularity as “a bunch of hot air,” and went on further to declare that “the brain is not computable and no engineering can reproduce it.”
Ray Kurzweil, no doubt, couldn’t disagree more. You know, the guy who’s last book was entitled “How To Create a Mind”?
But Nicolelis isn’t backing down from critics. A very lively Twitter discussion took place in the days after he made the comments. “How in heavens do you simulate something you have no algorithm for?” went one Tweet. “…we would not be talking about consciousness. Our brain is ‘copy-write’ protected by its own evolutionary history!” went another. And the most damning hurl in the direction of Singularitarians: “Fallacy is what people are selling: that human nature can be reduced to [something] that [a] computer algorithm can run! This is a new church!”
I had to use Google Translate to translate the comments from Portuguese, but they seem to be accurate translations given his argument at the meeting.
Describing his new Pattern Recognition Theory of Mind (PRTM) during a Singularity Hub drive along interview last October, Ray Kurzweil voiced an opinion that couldn’t be more different from Nicolelis’. “We now have enough evidence to support a particular theory, …a uniform theory about how the neocortex works. And it’s basically comprised of 300 million pattern recognizers. Most important they can wire themselves in hierarchies to other pattern recognizers. The world is inherently hierarchical and the neocortex allows us to understand it in that hierarchical fashion.”
That is to say, Kurzweil thinks there is a certain simplicity to the structure of the neocortex, the part of the brain where the most complex human mental activities take place, that lends itself to being reproduced – by 2029, he famously predicts.
But Nicolelis isn’t buying it. He thinks the brain/neocortex is much more than a hierarchy of pattern recognizers, and it’s that complexity that futurists like Kurzweil underestimate. “You can’t predict whether the stock market will go up or down because you can’t compute it,” he said at the AAAS conference, MIT Technology Review reports. “You could have all the computer chips ever in the world and you won’t create a consciousness.”
Even if he thinks human thought won’t one day be recreated with silicon, Nicolelis certainly believes it will be augmented by it. A leading researcher in brain-computer interface technologies, he presented such an apparatus at the AAAS meeting: a device that enabled rats to detect infrared light. They did this by mounting on the rats’ heads infrared sensors that were connected to stimulating electrodes implanted in the brain. Infrared signals were translated to a stimulation pattern in the area of the brain that processes touch sensations, the somatosensory cortex. In this way they created a “sensory neuroprosthesis.” He hopes that these sorts of devices may one day “serve to expand natural perceptual capabilities in mammals.”
Of course, Nicolelis is not the first to suggest Kurzweil and his Singularitarian followers could use more hard science and less wishful thinking in their predictions. In fact, New York University psychology professor Gary Marcus wrote a scathing condemnation of Kurzweil’s “How To Create A Mind” in the New Yorker: “Kurzweil’s pointers to neuroanatomy serve more as razzle-dazzle than real evidence for his theory” is the take-home message.
As he always does, Kurzweil is sure to fire back at his critics. And as the newly appointed Director of Engineering at Google, where his explicit mission is to create an artificial intelligence that will “make all of us smarter,” he’s certainly got the money to put where his mouth is. He’s teaming up with Google to create the most sophisticated AI assistant the world has ever seen. It might not be the brain that many of his fans are waiting for, but it could very well help the cause. Because it’s already 2013, and 2029 will be here before you know it.
EDIT: There is something eerie in the fact that Geth started tracking this post immediately from my posting it.
EDIT2: And the tracking disappears just as quickly
Though, as far as I'm concerned, I think it'd be a bit of a blessing if it turns out we can't infuse silicon with whatever "consciousness" is. Won't ever have to deal with the funky moral grey zones or anything, and we can all breathe easy again. The research in and of itself though will probably have several boons, intended as well as unintended ones, for the rest of us. There are certainly worse things to spend your money on.
Alright and in this next scene all the animals have AIDS.
The problem with that is, near as I understand it, "consciousness" isn't something you infuse things with, it's an emergent property of a sufficiently complex processing system.
So what is his actual reasoning behind our never being able to reproduce the brain?
The singularity is nonsense, but "strong AI is impossible" seems equally nonsensical.
Going by the article's translation of his tweets "It's too complex"
from the MIT Technology Review piece linked in the article:
But Nicolelis is in a camp that thinks that human consciousness (and if you believe in it, the soul) simply can’t be replicated in silicon. That’s because its most important features are the result of unpredictable, nonlinear interactions among billions of cells, Nicolelis says.
“You can’t predict whether the stock market will go up or down because you can’t compute it,” he says. “You could have all the computer chips ever in the world and you won’t create a consciousness.”
The neuroscientist, originally from Brazil, instead thinks that humans will increasingly subsume machines (an idea, incidentally, that’s also part of Kurzweil’s predictions).
I don't understand how otherwise brilliant people can so sternly stick to their beliefs on a subject on which we have no definite answer. You can say you think it is one way or another, sure, but the factual answer is that we have no idea whether we can fully recreate the brain because we do not know how it functions.
Grey Paladin on
"All men dream, but not equally. Those who dream by night in the dusty recesses of their minds wake in the day to find that it was vanity; but the dreamers of the day are dangerous men, for they may act their dream with open eyes to make it possible." - T.E. Lawrence
from the MIT Technology Review piece linked in the article:
But Nicolelis is in a camp that thinks that human consciousness (and if you believe in it, the soul) simply can’t be replicated in silicon. That’s because its most important features are the result of unpredictable, nonlinear interactions among billions of cells, Nicolelis says.
That's a computational problem. It's entirely possible to simulate nonlinear, nonbinary interactions of complex systems; given enough money and silicon and processing time. And "unpredictable" on a macro level (as opposed to a quantum level, and I really doubt that quantum interactions have a non-negligible effect on consciousness) just means "we don't understand it... yet."
I think that Kurzweil has no basis for his 2029 prediction. I'd get behind Nicolelis if he were saying, "It's probably not possible in our lifetimes," but saying that it's not possible at all ever is a bit silly.
every person who doesn't like an acquired taste always seems to think everyone who likes it is faking it. it should be an official fallacy.
I sort of find it amusing as well that Nicolelis is pushing neurological augmentation pretty hard while downplaying the ability to ever mimic the brain computationally. It's pretty interesting however:
From the same piece:
Similarly, Nicolelis thinks in the future humans with brain implants might be able to sense x-rays, operate distant machines, or navigate in virtual space with their thoughts, since the brain will accommodate foreign objects including computers as part of itself.
Recently, Nicolelis’s Duke lab has been looking to put an exclamation point on these ideas. In one recent experiment, they used a brain implant so that a monkey could control a full-body computer avatar, explore a virtual world, and even physically sense it.
In other words, the human brain creates models of tools and machines all the time, and brain implants will just extend that capability. Nicolelis jokes that if he ever opened a retail store for brain implants, he’d call it Machines “R” Us.
As best I can tell, Nicolelis is saying that even if you could manually reproduce every particle in a brain and whatever supporting systems are necessary, it still wouldn't be a consciousness. That's what I'm taking away from "no engineering can reproduce it". Which makes no sense unless you're assuming the existence of some noncorporeal element, something that - whether it's true or not - is bad science more or less by definition.
I submitted an entry to Lego Ideas, and if 10,000 people support me, it'll be turned into an actual Lego set!If you'd like to see and support my submission, follow this link.
The problem with that is, near as I understand it, "consciousness" isn't something you infuse things with, it's an emergent property of a sufficiently complex processing system.
Yeah, that was sloppy wording on my part, I should've said something like "giving silicon consciousness", but that sounds about right.
As best I can tell, Nicolelis is saying that even if you could manually reproduce every particle in a brain and whatever supporting systems are necessary, it still wouldn't be a consciousness. That's what I'm taking away from "no engineering can reproduce it". Which makes no sense unless you're assuming the existence of some noncorporeal element, something that - whether it's true or not - is bad science more or less by definition.
I think the most correct assessment, if you have to go as far as saying impossible, would be to say "Impossible to manufacture using nothing but silicon".
EDIT: And when I thin about it, that probably needs a "With our current manufacturing technology" qualifier as well.
Zephiran on
Alright and in this next scene all the animals have AIDS.
Brain cell interactions per se are already immenselly more complex than whatever we're able to dream, let alone the interactions between billions of those amazingly complex units. It's not likely that we can just add more CPUs to a monster machine until it gets as complex as a brain and thus develops awareness. Because the two systems are nothing alike in any possible sense.
I'm a fair-weather fan of the singularity, but I'm also open to the possibility that there is just something inherently non-binary / non-digital / non-symbolic about the brain and consciousness that can't possibly be reproduced completely by any binary/digital/symbolic instruction set.
That doesn't mean something non-physical like a soul. In fact, just the opposite. Anything created algorithmically means it is ultimately reduced to the non-physical. It is symbolic communication, not corporeal being. I don't know for sure that I believe that anything in the universe of reality can be perfectly represented by a symbolic communication. At some point the symbolic communication is only an abstract model of something "real" or else is a placeholder for "that which we still do not know." To know it is to raze the model, and thus the symmetry between algorithm and life.
Yeah, if you want to say something like "modern computer technology cannot perfectly model the brain" or even, like Yar suggests, "binary systems cannot perfectly model the brain," I can get on board that train. Not to say I necessarily agree with those sentiments, because I don't know nearly enough about any of this stuff to opine, but they're logically defensible. "Making a brain is an engineering impossibility" is just on par with "heavier-than-air flight is impossible" or "nobody will ever need more than 64k of memory."
As to perfect symbolic modeling of something in the universe, I'm pretty sure it's impossible, if nothing else, because of the uncertainty principle. I mean, it's fundamentally impossible to perfectly describe the state of any object, unless our understanding of quantum mechanics is seriously wrong. That doesn't mean, of course, that we can't represent something like a brain with sufficient accuracy that the errors are irrelevant, just as my inability to perfectly describe the state of a bullet doesn't prevent me from telling you where it will land with amazing precision.
I submitted an entry to Lego Ideas, and if 10,000 people support me, it'll be turned into an actual Lego set!If you'd like to see and support my submission, follow this link.
“We now have enough evidence to support a particular theory, …a uniform theory about how the neocortex works. And it’s basically comprised of 300 million pattern recognizers. Most important they can wire themselves in hierarchies to other pattern recognizers. ”
The problem with that is, near as I understand it, "consciousness" isn't something you infuse things with, it's an emergent property of a sufficiently complex processing system.
I wouldn't characterize consciousness (as poorly defined as the concept is) as something that simply comes about as from a complex system. In the same way I wouldn't characterize a pancreas as something that simply comes about from a complex enough digestive system.
Consciousness, like any other aspect of a biological system, fulfills or previously fulfilled some biological need that was selected for in the environment. Evolution brought about consciousness.
Just because a system is as complex as the human brain does not mean it will bring about consciousness.
I'm a fair-weather fan of the singularity, but I'm also open to the possibility that there is just something inherently non-binary / non-digital / non-symbolic about the brain and consciousness that can't possibly be reproduced completely by any binary/digital/symbolic instruction set.
Abstraction is key. I don't doubt that we will one day be able emulate an entire human brain on traditional computer architecture. Given enough computing power, there's no theoretical reason you couldn't emulate an entire brain down to the atomic level.
But I doubt the usefulness of such an endeavor. Allow me to quote text from “Whole Brain Emulation: A Roadmap”:
An important hypothesis for [Whole Brain Emulation] is that in order to emulate the brain we do not need to understand the whole system, but rather we just need a database containing all necessary low-level information about the brain and knowledge of the local update rules that change brain states from moment to moment. A functional understanding (why is a particular piece of cortex organized in a certain way) is logically separate from detail knowledge (how is it organized, and how does this structure respond to signals). Functional understanding may be a possible result from detail knowledge and it may help gather only the relevant information for WBE, but it is entirely possible that we could acquire full knowledge of the component parts and interactions of the brain without gaining an insight into how these produce (say) consciousness or intelligence.
When people unite together, they become stronger than the sum of their parts.
Don't assume bad intentions over neglect and misunderstanding.
I'm a fair-weather fan of the singularity, but I'm also open to the possibility that there is just something inherently non-binary / non-digital / non-symbolic about the brain and consciousness that can't possibly be reproduced completely by any binary/digital/symbolic instruction set.
I disagree - there's a reason for instance, that no non-binary computer has ever been created. Because any conceivable number system reduces itself to a binary instruction set, just of a particular size. This seems to scale all the way down to the smallest elements the brain could be influenced by, since at the fundamental level we know the brain can't measure things in Planck-time and probably can't measure things at much higher levels of fidelity (atomic position, molecular position, probably precise protein and DNA structures).
However "analog" a brain may appear to be, there's no reason to think it couldn't be reduced to a sufficiently long binary code that could be manipulated by a computer.
Of course to some degree all the detail is in "reproduced completely", because the nature of consciousness itself is strikingly incomplete and pretty highly fluid. If I get hit in the head sufficiently hard - which has happened, from time to time - then some degree of low-level neural damage is likely, and at the very least there's been a massive state change across my brain. But we generally don't think that people "die" when this happens, so to me it seems that a complete reproduction is unnecessary - the degree of replication for someone to retain their identity, seems more and more like it might be far less then a perfect replication of the entire brain.
from the MIT Technology Review piece linked in the article:
But Nicolelis is in a camp that thinks that human consciousness (and if you believe in it, the soul) simply can’t be replicated in silicon. That’s because its most important features are the result of unpredictable, nonlinear interactions among billions of cells, Nicolelis says.
That's a computational problem. It's entirely possible to simulate nonlinear, nonbinary interactions of complex systems; given enough money and silicon and processing time. And "unpredictable" on a macro level (as opposed to a quantum level, and I really doubt that quantum interactions have a non-negligible effect on consciousness) just means "we don't understand it... yet."
I think that Kurzweil has no basis for his 2029 prediction. I'd get behind Nicolelis if he were saying, "It's probably not possible in our lifetimes," but saying that it's not possible at all ever is a bit silly.
There are some theoretical limits to Turing Machines. For instance, Turing Machines are discrete state machines, and as such cannot necessarily perform the same calculations as continuous state machines (like differential analyzers). So it's not the case that enough money and silicon can get you literally anything in the computational realm. But, to be fair, it's doubtful that this particular difference has that much to do with consciousness or any of the other particularly interesting properties of the human mind.
edit: I would say the same re ELM above. I'm going off Turing's authority, which may have been superseded in the last half century, but he seemed convinced that there were computational tasks that could not be completed by a digital computer.
Even if the phenomenon of consciousness couldn't be reproduced with binary information storage/processing, it's still a physically reducible system. Any physically reducible system should be reproducable as well, though this may occur by accident or through a non-deliberate artificial selection process or what have you.
I'm a fair-weather fan of the singularity, but I'm also open to the possibility that there is just something inherently non-binary / non-digital / non-symbolic about the brain and consciousness that can't possibly be reproduced completely by any binary/digital/symbolic instruction set.
I disagree - there's a reason for instance, that no non-binary computer has ever been created. Because any conceivable number system reduces itself to a binary instruction set, just of a particular size. This seems to scale all the way down to the smallest elements the brain could be influenced by, since at the fundamental level we know the brain can't measure things in Planck-time and probably can't measure things at much higher levels of fidelity (atomic position, molecular position, probably precise protein and DNA structures).
However "analog" a brain may appear to be, there's no reason to think it couldn't be reduced to a sufficiently long binary code that could be manipulated by a computer.
Of course to some degree all the detail is in "reproduced completely", because the nature of consciousness itself is strikingly incomplete and pretty highly fluid. If I get hit in the head sufficiently hard - which has happened, from time to time - then some degree of low-level neural damage is likely, and at the very least there's been a massive state change across my brain. But we generally don't think that people "die" when this happens, so to me it seems that a complete reproduction is unnecessary - the degree of replication for someone to retain their identity, seems more and more like it might be far less then a perfect replication of the entire brain.
This might seem pedantic, but binary isn't the only option...
While I can get behind the statements that 'no modern computer can reproduce the human brain', or 'we won't see a computer simulate a brain in our lifetime'. Hell, I'd give a grudging 'maybe' to a statement like 'human consciousness is intractably related to our senses and experiences, and can't be directly recreated on a machine'.
But some form of consciousness? I think it's quite possible. We know that a brain can be created - everyone on earth is born with one. I don't think there are any engineering problems that are entirely intractable or impossible. At some point, we'll potentially have machines that are powerful enough to simulate a real human brain with arbitrary precision. I think it's an open question the role things like quantum mechanics and uncertainty effects play in conscious thought, and the chaotic affect they will have may make it difficult to simulate real and ongoing human thought.
When it comes down to it though, I am pretty confident that we'll find consciousness - at least some form of it - is an emergent property of any sufficiently complex processing system.
from the MIT Technology Review piece linked in the article:
But Nicolelis is in a camp that thinks that human consciousness (and if you believe in it, the soul) simply can’t be replicated in silicon. That’s because its most important features are the result of unpredictable, nonlinear interactions among billions of cells, Nicolelis says.
That's a computational problem. It's entirely possible to simulate nonlinear, nonbinary interactions of complex systems; given enough money and silicon and processing time. And "unpredictable" on a macro level (as opposed to a quantum level, and I really doubt that quantum interactions have a non-negligible effect on consciousness) just means "we don't understand it... yet."
I think that Kurzweil has no basis for his 2029 prediction. I'd get behind Nicolelis if he were saying, "It's probably not possible in our lifetimes," but saying that it's not possible at all ever is a bit silly.
There are some theoretical limits to Turing Machines. For instance, Turing Machines are discrete state machines, and as such cannot necessarily perform the same calculations as continuous state machines (like differential analyzers). So it's not the case that enough money and silicon can get you literally anything in the computational realm. But, to be fair, it's doubtful that this particular difference has that much to do with consciousness or any of the other particularly interesting properties of the human mind.
edit: I would say the same re ELM above. I'm going off Turing's authority, which may have been superseded in the last half century, but he seemed convinced that there were computational tasks that could not be completed by a digital computer.
I don't think it was an opinion, I think you can mathematically prove that a computer cannot perform any operation. Even if it isn't a proof, I would think computers would still be limited on account of Godel. Of course, I don't see why human brains can't be limited in a similar way.
I really think the problem with the brain is that it's very reliant on constant input from the body. I even read somewhere that the basis for our self-consciousness was actually the constant input from our internal sensors. So your brain "knows" that "you are you" because there is a plethora of signals coming from your gut, heart, lungs etc. that make it able to identify with a body and keep track of temporal processes, a frame of reference, maybe.
Of course THAT could also be simulated.
I can't really give an informed opinion on whether singularity is possible or not. I'm just going to say that I work in auditory neuroscience and that people who make cochlear implants and speech recognizers, which are incredibly simple stuff compared with what a brain simulation would require, constantly run into trouble simulating these relatively simple biological processes and can 't match the level of performance that our brain (or a rat's, for that matter) have. So I think that Kurzweil is overly optimistic.
Also, whole cortical layers full of auditory neurons can change their distribution of preferred frequencies at the drop of a hat, and I guess it's the same for the rest of the brain, so this is all much more complex than just modelling every single neuron. Brains are tetradimensional machines.
from the MIT Technology Review piece linked in the article:
But Nicolelis is in a camp that thinks that human consciousness (and if you believe in it, the soul) simply can’t be replicated in silicon. That’s because its most important features are the result of unpredictable, nonlinear interactions among billions of cells, Nicolelis says.
That's a computational problem. It's entirely possible to simulate nonlinear, nonbinary interactions of complex systems; given enough money and silicon and processing time. And "unpredictable" on a macro level (as opposed to a quantum level, and I really doubt that quantum interactions have a non-negligible effect on consciousness) just means "we don't understand it... yet."
I think that Kurzweil has no basis for his 2029 prediction. I'd get behind Nicolelis if he were saying, "It's probably not possible in our lifetimes," but saying that it's not possible at all ever is a bit silly.
There are some theoretical limits to Turing Machines. For instance, Turing Machines are discrete state machines, and as such cannot necessarily perform the same calculations as continuous state machines (like differential analyzers). So it's not the case that enough money and silicon can get you literally anything in the computational realm. But, to be fair, it's doubtful that this particular difference has that much to do with consciousness or any of the other particularly interesting properties of the human mind.
edit: I would say the same re ELM above. I'm going off Turing's authority, which may have been superseded in the last half century, but he seemed convinced that there were computational tasks that could not be completed by a digital computer.
I don't think it was an opinion, I think you can mathematically prove that a computer cannot perform any operation. Even if it isn't a proof, I would think computers would still be limited on account of Godel. Of course, I don't see why human brains can't be limited in a similar way.
The only task I'm aware of is the halting problem - a computer can't predict when a given program will cease running until it has run to completion.
But that's a very general and fundamental problem, the general solution to which would solve all manner of currently unsolvable mathematical paradoxes - but it's not a commentary on the human brain, which cannot do these things either nor operates in a manner which requires a solution.
The discrete vs continuous state is what I was alluding to with the comment on Planck length. The universe is not apparently continuously measurable, so we know there are some significant lower bounds on how small things can be and still matter, but it seems more significant to me that the brain doesn't appear to engage in anything which requires a continuous process - human perception is only apparently continuous, but we know that neural communication itself works in a binary-like manner with electrical pulses, and not continually varying potentials.
from the MIT Technology Review piece linked in the article:
But Nicolelis is in a camp that thinks that human consciousness (and if you believe in it, the soul) simply can’t be replicated in silicon. That’s because its most important features are the result of unpredictable, nonlinear interactions among billions of cells, Nicolelis says.
That's a computational problem. It's entirely possible to simulate nonlinear, nonbinary interactions of complex systems; given enough money and silicon and processing time. And "unpredictable" on a macro level (as opposed to a quantum level, and I really doubt that quantum interactions have a non-negligible effect on consciousness) just means "we don't understand it... yet."
I think that Kurzweil has no basis for his 2029 prediction. I'd get behind Nicolelis if he were saying, "It's probably not possible in our lifetimes," but saying that it's not possible at all ever is a bit silly.
There are some theoretical limits to Turing Machines. For instance, Turing Machines are discrete state machines, and as such cannot necessarily perform the same calculations as continuous state machines (like differential analyzers). So it's not the case that enough money and silicon can get you literally anything in the computational realm. But, to be fair, it's doubtful that this particular difference has that much to do with consciousness or any of the other particularly interesting properties of the human mind.
edit: I would say the same re ELM above. I'm going off Turing's authority, which may have been superseded in the last half century, but he seemed convinced that there were computational tasks that could not be completed by a digital computer.
I don't think it was an opinion, I think you can mathematically prove that a computer cannot perform any operation. Even if it isn't a proof, I would think computers would still be limited on account of Godel. Of course, I don't see why human brains can't be limited in a similar way.
The only task I'm aware of is the halting problem - a computer can't predict when a given program will cease running until it has run to completion.
But that's a very general and fundamental problem, the general solution to which would solve all manner of currently unsolvable mathematical paradoxes - but it's not a commentary on the human brain, which cannot do these things either nor operates in a manner which requires a solution.
The discrete vs continuous state is what I was alluding to with the comment on Planck length. The universe is not apparently continuously measurable, so we know there are some significant lower bounds on how small things can be and still matter, but it seems more significant to me that the brain doesn't appear to engage in anything which requires a continuous process - human perception is only apparently continuous, but we know that neural communication itself works in a binary-like manner with electrical pulses, and not continually varying potentials.
Just a brief aside here: maybe neurons aren't all thats going on in brain operation. Glial cells, previously known mostly as structural cells, have been shown to influence the intelligence of mice.
Today we are closer than ever to understanding the biological basis of human thought. In a major first for neuroscience, researchers have produced an image showing almost an entire vertebrate brain at work -- down to the level of individual neurons. Soon we'll have a human brain "activity map" which reveals how electrical impulses in the brain correlate to thought patterns, biological processes, and more.
The neurons in question belong to a zebrafish embryo, and the researchers come from HHMI's Janelia Farm Research Campus. In the video up top, the activity of individual neurons appear as flashes, detonating across the fish's entire larval brain. And while the brain of a zebrafish only contains about 100,000 neurons (compared to the tens of billions in the human brain), it represents an important step along the path to creating a Brain Activity Map for us apes, a project into which the Obama administration may soon funnel billions of dollars.
According to findings published in the latest issue of Nature Methods, microscopist Phillip Keller and neurobiologist Misha Ahrens have modified an existing imaging technique (called light sheet microscopy) in such a way that enables them to record neuronal activity from the entire volume of the zebrafish's brain. They did this while the embryo was alive, and with a temporal resolution of 0.8 Hz (meaning they were recording activity about once every second). All told, Keller and Ahrens were able to capture "more than 80% of all neurons at single-cell resolution. "
Emphasis added, because that last bit is important. Imaging whole-brain activity isn't really new. Neither is mesauring the activity of single neurons. Doing both simultaneously, however, is really impressive, and hugely valuable from an experimental standpoint -- akin being able to see the individual dots of a pointillist masterpiece and the painting as a whole all at once.
The researchers used their newfound imaging abilities to demonstrate how their novel technique could be used to uncover two functionally defined neuron-circuits across the zebrafish's brain. The video below is a 3D visualization of a circuit of neurons extending from the zebrafish's hindbrain into its spinal cord.
“It’s phenomenal,” says Yuste... “It is a bright star now in the literature, suggesting that it is not crazy to map every neuron in the brain of an animal.” Yuste has been leading the call for a big biology project that would do just that in the human brain, which contains about 85,000 times more neurons than the zebrafish brain.
The resolution offered by the zebrafish study will enable researchers to understand how different regions of the brain work together, says Ahrens. With conventional techniques, imaging even 2,000 neurons at once is difficult, so researchers must pick and choose which to look at, and extrapolate. Now, he says, “you don't need to guess what is happening — you can see it”.
It'll be interesting to see what impact existing brain-mapping projects like this one have on bringing Yuste's proposed project to fruition. Many scientists have complained that such an undertaking lacks the clear conceptual aims of other major, government-funded research endeavors (the Human Genome Project, for example).
Others have voiced concern that a massive Brain Mapping Project would lead to the messy reapportioning of already limited scientific funding. Will studies like this one, which demonstrate the feasibility of whole brain mapping in less complex organisms, lead some of them to reconsider their grievances? Or will its shortcomings (the fact that it really only works in transparent organisms -- like zebrafish embryos) keep it from being taken into serious consideration?
The researchers' findings are published in the latest issue of Nature Methods (no paywall!). Check out a detailed account of the work at Nature News.
I think that dude is opposing brain simulation on empirical or analytical science grounds. The best laid, logically sound, plans can go terribly awry in the pragmatic and empirical bench environment.
0
Options
surrealitychecklonely, but not unloveddreaming of faulty keys and latchesRegistered Userregular
also
"we cant simulate the stock market" does not equal "we cannot simulate a stock market that satisfies our stock market criteria"
"we cant simulate the stock market" does not equal "we cannot simulate a stock market that satisfies our stock market criteria"
Also since simulating the stock market accurately would in turn require, possibly, the simulation of the brain's of millions of stock market actors itself.
It's a poor litmus test.
0
Options
ShivahnUnaware of her barrel shifter privilegeWestern coastal temptressRegistered User, Moderatormod
Um, I might be totally misunderstanding the meaning of the word analog, but... why are we treating brains as analog devices? The computational parts are all neurons that are necessarily, by design, either firing or not firing. There is no in between state, at any given point in time, each synapse's presynaptic cell is either firing or not firing.
I realize that pretty much everyone disagrees with me and/or has not brought this up, so I'm... assuming I'm wrong, but could someone explain how or why? Because I don't know how different the brain is from a large parallel structure of binary devices operating at 100 Hertz.
Earlier in this thread, we started talking about life for some reason, and I guess I just picked up on it now. I want to note that life has a somewhat weak relevance to intelligence. Just ask your friendly neighbourhood dandelion.
Maybe it's a bit human-centric, too. The thing with reproducing human brains.
Um, I might be totally misunderstanding the meaning of the word analog, but... why are we treating brains as analog devices? The computational parts are all neurons that are necessarily, by design, either firing or not firing. There is no in between state, at any given point in time, each synapse's presynaptic cell is either firing or not firing.
I realize that pretty much everyone disagrees with me and/or has not brought this up, so I'm... assuming I'm wrong, but could someone explain how or why? Because I don't know how different the brain is from a large parallel structure of binary devices operating at 100 Hertz.
The inputs that determine if the neuron are firing are electrochemical charges and quantities of neurotransmitters that are binding to receptors, these are functionally analog.
I'm also not 100% on the firing or not firing thing you just said. I mean, there are plenty of chemicals floating around in the brain that that partially inhibit or encourage ion transfer between neurons.
Um, I might be totally misunderstanding the meaning of the word analog, but... why are we treating brains as analog devices? The computational parts are all neurons that are necessarily, by design, either firing or not firing. There is no in between state, at any given point in time, each synapse's presynaptic cell is either firing or not firing.
I realize that pretty much everyone disagrees with me and/or has not brought this up, so I'm... assuming I'm wrong, but could someone explain how or why? Because I don't know how different the brain is from a large parallel structure of binary devices operating at 100 Hertz.
The inputs that determine if the neuron are firing are electrochemical charges and quantities of neurotransmitters that are binding to receptors, these are functionally analog.
I'm also not 100% on the firing or not firing thing you just said. I mean, there are plenty of chemicals floating around in the brain that that partially inhibit or encourage ion transfer between neurons.
But they do respond to receptors, and receptors are frequently binary-like, even if there's fair quantities of them. Neurons generally have action-potentials which allow for charge-accumulation type activities - that's not so much analog as more like an A2D input and it's definitely not infinite-resolution.
Posts
Best sentence.
Neuroscience is interesting because it's on the cusp of a shift. For a long time, we knew very little about the brain. Once study became feasible, we did what we could. But the field is still in its childhood. There is lots of exploratory research to do still - just, "holy shit what does this do?" Glial cells have not been studied nearly enough. I saw a talk last month about... astrocytes or microglia (oh god I'm the worst) doing some stuff we had never even thought of. So there are still total mysteries to be probed and we don't know what questions to ask.
However, similar to physics in the beginning of the twentieth century, we're hitting a point where we're going to start seeing more modeling. In a mature field, most of what is done is trying to find better models for most phenomena, as opposed to just poking things because we don't know what happens when we poke them. Neuroscience is starting to see that - more and more models to reflect the underlying world, even while new discoveries about things we never knew we never knew are made.
So the field is in an interesting position. It's straddling the line between a totally new and mature discipline. Essentially, it's in its adolescence, where it's just starting to become very mature and useful while still having that childish wonder about it.
http://singularityhub.com/2013/03/10/leading-neuroscientist-says-kurzweil-singularity-prediction-a-bunch-of-hot-air/
EDIT: There is something eerie in the fact that Geth started tracking this post immediately from my posting it.
EDIT2: And the tracking disappears just as quickly
And Brain rats are scary. Torment flashbacks.
Though, as far as I'm concerned, I think it'd be a bit of a blessing if it turns out we can't infuse silicon with whatever "consciousness" is. Won't ever have to deal with the funky moral grey zones or anything, and we can all breathe easy again. The research in and of itself though will probably have several boons, intended as well as unintended ones, for the rest of us. There are certainly worse things to spend your money on.
I got a little excited when I saw your ship.
The problem with that is, near as I understand it, "consciousness" isn't something you infuse things with, it's an emergent property of a sufficiently complex processing system.
The singularity is nonsense, but "strong AI is impossible" seems equally nonsensical.
Going by the article's translation of his tweets "It's too complex"
from the MIT Technology Review piece linked in the article:
http://www.technologyreview.com/view/511421/the-brain-is-not-computable/
That's a computational problem. It's entirely possible to simulate nonlinear, nonbinary interactions of complex systems; given enough money and silicon and processing time. And "unpredictable" on a macro level (as opposed to a quantum level, and I really doubt that quantum interactions have a non-negligible effect on consciousness) just means "we don't understand it... yet."
I think that Kurzweil has no basis for his 2029 prediction. I'd get behind Nicolelis if he were saying, "It's probably not possible in our lifetimes," but saying that it's not possible at all ever is a bit silly.
the "no true scotch man" fallacy.
From the same piece:
Yeah, that was sloppy wording on my part, I should've said something like "giving silicon consciousness", but that sounds about right.
I think the most correct assessment, if you have to go as far as saying impossible, would be to say "Impossible to manufacture using nothing but silicon".
EDIT: And when I thin about it, that probably needs a "With our current manufacturing technology" qualifier as well.
I got a little excited when I saw your ship.
That doesn't mean something non-physical like a soul. In fact, just the opposite. Anything created algorithmically means it is ultimately reduced to the non-physical. It is symbolic communication, not corporeal being. I don't know for sure that I believe that anything in the universe of reality can be perfectly represented by a symbolic communication. At some point the symbolic communication is only an abstract model of something "real" or else is a placeholder for "that which we still do not know." To know it is to raze the model, and thus the symmetry between algorithm and life.
Does Kurzwiel's idea of the Singularity require artificial intelligence? Either as an invention or an emergent property?
As to perfect symbolic modeling of something in the universe, I'm pretty sure it's impossible, if nothing else, because of the uncertainty principle. I mean, it's fundamentally impossible to perfectly describe the state of any object, unless our understanding of quantum mechanics is seriously wrong. That doesn't mean, of course, that we can't represent something like a brain with sufficient accuracy that the errors are irrelevant, just as my inability to perfectly describe the state of a bullet doesn't prevent me from telling you where it will land with amazing precision.
Can't we already do this somewhat reliably with, you know, a sophisticated collection of water tanks?
What is Kurzweil referring to here?
I wouldn't characterize consciousness (as poorly defined as the concept is) as something that simply comes about as from a complex system. In the same way I wouldn't characterize a pancreas as something that simply comes about from a complex enough digestive system.
Consciousness, like any other aspect of a biological system, fulfills or previously fulfilled some biological need that was selected for in the environment. Evolution brought about consciousness.
Just because a system is as complex as the human brain does not mean it will bring about consciousness.
Abstraction is key. I don't doubt that we will one day be able emulate an entire human brain on traditional computer architecture. Given enough computing power, there's no theoretical reason you couldn't emulate an entire brain down to the atomic level.
But I doubt the usefulness of such an endeavor. Allow me to quote text from “Whole Brain Emulation: A Roadmap”:
Don't assume bad intentions over neglect and misunderstanding.
I disagree - there's a reason for instance, that no non-binary computer has ever been created. Because any conceivable number system reduces itself to a binary instruction set, just of a particular size. This seems to scale all the way down to the smallest elements the brain could be influenced by, since at the fundamental level we know the brain can't measure things in Planck-time and probably can't measure things at much higher levels of fidelity (atomic position, molecular position, probably precise protein and DNA structures).
However "analog" a brain may appear to be, there's no reason to think it couldn't be reduced to a sufficiently long binary code that could be manipulated by a computer.
Of course to some degree all the detail is in "reproduced completely", because the nature of consciousness itself is strikingly incomplete and pretty highly fluid. If I get hit in the head sufficiently hard - which has happened, from time to time - then some degree of low-level neural damage is likely, and at the very least there's been a massive state change across my brain. But we generally don't think that people "die" when this happens, so to me it seems that a complete reproduction is unnecessary - the degree of replication for someone to retain their identity, seems more and more like it might be far less then a perfect replication of the entire brain.
There are some theoretical limits to Turing Machines. For instance, Turing Machines are discrete state machines, and as such cannot necessarily perform the same calculations as continuous state machines (like differential analyzers). So it's not the case that enough money and silicon can get you literally anything in the computational realm. But, to be fair, it's doubtful that this particular difference has that much to do with consciousness or any of the other particularly interesting properties of the human mind.
edit: I would say the same re ELM above. I'm going off Turing's authority, which may have been superseded in the last half century, but he seemed convinced that there were computational tasks that could not be completed by a digital computer.
This might seem pedantic, but binary isn't the only option...
Ternary computer
Analog computer
While I can get behind the statements that 'no modern computer can reproduce the human brain', or 'we won't see a computer simulate a brain in our lifetime'. Hell, I'd give a grudging 'maybe' to a statement like 'human consciousness is intractably related to our senses and experiences, and can't be directly recreated on a machine'.
But some form of consciousness? I think it's quite possible. We know that a brain can be created - everyone on earth is born with one. I don't think there are any engineering problems that are entirely intractable or impossible. At some point, we'll potentially have machines that are powerful enough to simulate a real human brain with arbitrary precision. I think it's an open question the role things like quantum mechanics and uncertainty effects play in conscious thought, and the chaotic affect they will have may make it difficult to simulate real and ongoing human thought.
When it comes down to it though, I am pretty confident that we'll find consciousness - at least some form of it - is an emergent property of any sufficiently complex processing system.
I don't think it was an opinion, I think you can mathematically prove that a computer cannot perform any operation. Even if it isn't a proof, I would think computers would still be limited on account of Godel. Of course, I don't see why human brains can't be limited in a similar way.
Complexity may be necessary for consciousness, but who is to say that it's sufficient?
Isn't he saying something like that? That the brain is a physical object, and a simulation will only be a simulation?
Of course THAT could also be simulated.
I can't really give an informed opinion on whether singularity is possible or not. I'm just going to say that I work in auditory neuroscience and that people who make cochlear implants and speech recognizers, which are incredibly simple stuff compared with what a brain simulation would require, constantly run into trouble simulating these relatively simple biological processes and can 't match the level of performance that our brain (or a rat's, for that matter) have. So I think that Kurzweil is overly optimistic.
Also, whole cortical layers full of auditory neurons can change their distribution of preferred frequencies at the drop of a hat, and I guess it's the same for the rest of the brain, so this is all much more complex than just modelling every single neuron. Brains are tetradimensional machines.
The only task I'm aware of is the halting problem - a computer can't predict when a given program will cease running until it has run to completion.
But that's a very general and fundamental problem, the general solution to which would solve all manner of currently unsolvable mathematical paradoxes - but it's not a commentary on the human brain, which cannot do these things either nor operates in a manner which requires a solution.
The discrete vs continuous state is what I was alluding to with the comment on Planck length. The universe is not apparently continuously measurable, so we know there are some significant lower bounds on how small things can be and still matter, but it seems more significant to me that the brain doesn't appear to engage in anything which requires a continuous process - human perception is only apparently continuous, but we know that neural communication itself works in a binary-like manner with electrical pulses, and not continually varying potentials.
Are walking and thinking sufficiently similar for that comparison?
Just a brief aside here: maybe neurons aren't all thats going on in brain operation. Glial cells, previously known mostly as structural cells, have been shown to influence the intelligence of mice.
http://blogs.scientificamerican.com/guest-blog/2013/03/07/human-brain-cells-make-mice-smart/
http://io9.com/this-image-could-be-the-first-step-toward-mapping-human-456366486
"we cant simulate the stock market" does not equal "we cannot simulate a stock market that satisfies our stock market criteria"
Also since simulating the stock market accurately would in turn require, possibly, the simulation of the brain's of millions of stock market actors itself.
It's a poor litmus test.
I realize that pretty much everyone disagrees with me and/or has not brought this up, so I'm... assuming I'm wrong, but could someone explain how or why? Because I don't know how different the brain is from a large parallel structure of binary devices operating at 100 Hertz.
Maybe it's a bit human-centric, too. The thing with reproducing human brains.
The inputs that determine if the neuron are firing are electrochemical charges and quantities of neurotransmitters that are binding to receptors, these are functionally analog.
I'm also not 100% on the firing or not firing thing you just said. I mean, there are plenty of chemicals floating around in the brain that that partially inhibit or encourage ion transfer between neurons.
But they do respond to receptors, and receptors are frequently binary-like, even if there's fair quantities of them. Neurons generally have action-potentials which allow for charge-accumulation type activities - that's not so much analog as more like an A2D input and it's definitely not infinite-resolution.