6 ms·

Simulated brain closer to thought

62 points by alexandros 18y ago

quoderat 18y ago
So many things are converging at once on the AI front. I think it's probably likely we will have human-level AI in my lifetime (I'm 32). Of course, AI, like nuclear fusion, is perpetually 50 years away, but this time I believe it's different.
- chadmalik 18y ago
  I really doubt digital computation will ever be a platform for intelligence as we understand it. Biological systems really are different than digital ones.
  - eru 18y ago
    I guess after we have created and understood intelligence once, we may be able to re-create it on much weaker hardware the second time. Perhaps one day we'll even be able to e.g. beat the Turing Test on standard 2009 hardware with enough algorithmic knowledge.
alexandros 18y ago
The understatement of the applications by the scientists makes me suspect that (a) they are being politically wise keeping it on the down low or (b) that the article's claims are exaggerated. If (a) is the case though, the potential consequences just blow my mind.. I mean, Robin Hanson has been talking about emulated minds on chips for a while, but I didn't think we were so close to real results. What this would do to arguments about religion, bioethics, irreducibles will make for very interesting discussion.
- mixmax 18y ago
  And very interesting science.
- swombat 18y ago
  "Mind-blowing" is the word indeed. As I read through this article, my first thought was, what if they do create a full, human-like brain. Would it be immoral to then shut it down?
  - icey 18y ago
    From a morality standpoint, what is the difference between shutting it down and an indefinite state of stasis? The brain wouldn't know any different, it would just cease to process for a period of time, and then resume processing later. It would be just like a blackout. Except since the brain only receives stimulus from a simulated environment, the environment could be frozen as well. So, if everything freezes and restarts at a later time, the brain would not perceive any missing time.
    - staticshock 18y ago
      You're assuming isolation. No feedback. It wouldn't perceive any missing time if it had no contact with the outer world. I think that's treating the brain as if it was just a computer.
      - icey 18y ago
        Ahh, that's an excellent point; for any of my argument to stand, it would definitely require full isolation (akin to what the current Blue Brain has right now).
  - inconvenient 18y ago
    If they simulate my brain molecule by molecule, and consciousness emerges, which one is really me? If my physical self is then killed, do I in fact still exist inside the simulation? In which case, is it OK to kill off my superfluous physical presence in order to use my organs to help others? What rights does the simulated me have? This will be a huge can of worms one day.
    - stcredzero 18y ago
- gruseom 18y ago
  I'm surprised that you thought the researchers were making understated claims: When asked when the simulation would come up with something artistic or an invention, Professor Markram said it was simply a matter of money. "[...T]he technology is there today." Boy does that sound like a lot of software projects I know.
- pygy 18y ago
  I think we're getting closer to AI, but I'm not as optimistic as the Blue Brain lead quote in the article. They are criticized for taking a reductionistic approach, and I actually agree with the critics. Metaphorically, they are now able to build puzzle pieces, but haven't yet figured how to make them fit, let alone how to assemble the big picture. Embryogenesis and development are very complicated processes, and, even though we also have a good idea of the macroscopic function of the cortex (primary, secondary areas for different tasks, associative areas ), some structure are still very mysterious and entangled. The brainstem is still a maze, for example, as is the so-called "mesoscopic" achitecture of the cortex.
newsit 18y ago
Sounds strangely familiar: http://www.simulation-argument.com/simulation.html http://www.simulation-argument.com/simulation.html
lucraft 18y ago
Seems unlikely that the researchers are really simulating this on a molecular level as claimed in that article. Perhaps they are using the neuron as the unit instead?
- pygy 18y ago
  Minsky first wired a virtual neural net in the 1950's. Some science got done meanwhile ;-) http://bluebrain.epfl.ch/page18924.html#2 http://bluebrain.epfl.ch/page18924.html#2
- paulgb 18y ago
  Sounds like you are right: "At the push of a button, the model could reconstruct biologically accurate neurons based on detailed experimental data, and automatically connect them in a biological manner, a task that involves positioning around 30 million synapses in precise 3D locations." [1] It doesn't necessarily mean they aren't modeling molecules, but the wording hints that they aren't. [1] http://bluebrain.epfl.ch/ http://bluebrain.epfl.ch/
Allocator2008 18y ago
One can only hope that as the reductionist principles upon which our consciousness operates are made apparent by projects like these, that the masses will finally begin to be disabused of the ever-persistent and pernicious god delusion which has impeded human progress and welfare since the dawn of recorded history.
- mcslee 18y ago
  Playing Devil's Advocate: if these simulations are truly modeled after humans, why are you so sure that they won't buy into many of the same religious ideas and reaffirm what many humans naturally think and do? Would a simulation like this even be aware that it was not just another person? It seems like the answer to this question MUST be that the simulation cannot tell the difference between itself and a normal human. Otherwise, your argument is moot, because the religious folk will simply be able to point at this thing and say "clearly this thing is different from us -- it even knows that it's different, so obviously this is not the way our 'souls' work."
  - paulgb 18y ago
    I think what Allocator2008 is getting at is that creating artificial life would show that there is nothing mystical or non-physical about consciousness. This would partly poke a hole in the argument for existence of God that the creation of life and/or consciousness has never been witnessed. To have the artificial brain not be aware it is not a regular person, or to have the brain buy into religious ideas, would actually make the point (that the artificial brain is just like a human's, and thus consciousness is not mystical or non-physical) stronger. Edit: khafra, ericb, and alexandros made the point I was trying to make better and in less words in replies to this thread: http://news.ycombinator.com/item?id=574258 http://news.ycombinator.com/item?id=574258
    - ewjordan 18y ago
      I think what Allocator2008 is getting at is that creating artificial life would show that there is nothing mystical or non-physical about consciousness. This would partly poke a hole in the argument for existence of God that the creation of life and/or consciousness has never been witnessed. Fear not. Even if the thing achieves consciousness, writes poetry, falls in love, critiques literature, etc., people will still claim that it's just faking it. "Neat trick, but it still doesn't have a soul, so it can't be conscious!" If I had to guess, I'd say that it would take a fresh generation, growing up with this sort of thing all around it, to accept (en masse) that the computers had whatever that special "something" is that we have. Personally, the chaos I'm really looking forward to is the legal mess that would arise - what happens when a computer, aided and guided by nobody in particular, does something illegal? Creates a piece of art and wants copyright protection? Who do you sue when a computer sets up its own Pirate Bay? Do you let it get its own bank account? What about when it figures out how to beat the stock market by hacking e-mail accounts of executives and returning trades based on insider information to clueless traders that just follow its black-box suggestions - is that insider trading if the system wasn't explicitly told to do that and no human ever sees the information? And so on...our legal system is neither equipped for this stuff, nor agile enough to adapt to it in a timely fashion, and these issues will remain unresolved far past the point of no return.
sown 18y ago
We are truly living in the future now.
- sown 18y ago
  Down-mod me if you want, but this kind of thing excites me! If there is magic in the world, then this is the kind I would expect to find in the future and I find it gratifying that we might understand it someday. :)
Darmani 18y ago
Much more promising to me than simulating larger regions of the brain is the ability to trace the execution of this region from the brain in detail and try to come up with a clue at how intelligence works. Evolution may be elegant itself, but it's not known for producing elegant systems. If either a human scientist or an AI is going to have any hope of designing efficient intelligent systems, we'll want something easier to work with than neurons.
jackchristopher 18y ago
This wouldn't prove consciousness is physical. I'm not saying I'm a dualist, but pragmatically, the only thing I can say is that I'm conscious. It doesn't matter if scientists showed me a conscious computer, that event is still happening in my consciousness. Why believe a computer the computer is conscious, when I can't even prove other humans are conscious? I don't necessarily believe the above, but no one mentioned this point. Also, the multiverse makes this an odder issue. If the simplest algorithm is to output all possible universes, what's stopping a universe from having just one observer?
- alexandros 18y ago
  Your position is dangerously close to solipsism. You allude to Occam's razor but it also answers your question. If the simplest algorithm does not imply limiting the amount of 'stuff' (universes), why would it limit the amount of observers?
  - jackchristopher 18y ago
    In principle I can't prove which I'm in. You're right. I left that out. But I see no reason to believe one over the other. But perpetual doubt is disempowering too. And when I face an undecidable with equally plausible answers, I'm forced to test each view to see the results: what's life like with view x vs. view y. That result becomes the metric for a view's worth. And honestly that could be solipsism. Although I know it as a similar concept some call "subjective reality". Again I don't believe in solipsism or SR, I'm just pointing out that either could be true and you could never know. I do believe in the multiverse though. I always thought computers as conscious would prove consciousness was physical. I didn't realize my mistake till recently.
- gambling8nt 18y ago
  Statistically, wouldn't one expect to find oneself in a many-observer universe (assuming that those universes weren't vastly rarer than the single-observer ones)?

62 points by alexandros 18y ago

So many things are converging at once on the AI front. I think it's probably likely we will have human-level AI in my lifetime (I'm 32). Of course, AI, like nuclear fusion, is perpetually 50 years away, but this time I believe it's different.

chadmalik 18y ago
I really doubt digital computation will ever be a platform for intelligence as we understand it. Biological systems really are different than digital ones.
- eru 18y ago
  I guess after we have created and understood intelligence once, we may be able to re-create it on much weaker hardware the second time. Perhaps one day we'll even be able to e.g. beat the Turing Test on standard 2009 hardware with enough algorithmic knowledge.

alexandros 18y ago

The understatement of the applications by the scientists makes me suspect that (a) they are being politically wise keeping it on the down low or (b) that the article's claims are exaggerated. If (a) is the case though, the potential consequences just blow my mind.. I mean, Robin Hanson has been talking about emulated minds on chips for a while, but I didn't think we were so close to real results. What this would do to arguments about religion, bioethics, irreducibles will make for very interesting discussion.

mixmax 18y ago
And very interesting science.
swombat 18y ago
"Mind-blowing" is the word indeed. As I read through this article, my first thought was, what if they do create a full, human-like brain. Would it be immoral to then shut it down?
- icey 18y ago
  From a morality standpoint, what is the difference between shutting it down and an indefinite state of stasis? The brain wouldn't know any different, it would just cease to process for a period of time, and then resume processing later. It would be just like a blackout. Except since the brain only receives stimulus from a simulated environment, the environment could be frozen as well. So, if everything freezes and restarts at a later time, the brain would not perceive any missing time.
  - staticshock 18y ago
    You're assuming isolation. No feedback. It wouldn't perceive any missing time if it had no contact with the outer world. I think that's treating the brain as if it was just a computer.
    - icey 18y ago
      Ahh, that's an excellent point; for any of my argument to stand, it would definitely require full isolation (akin to what the current Blue Brain has right now).
- inconvenient 18y ago
  If they simulate my brain molecule by molecule, and consciousness emerges, which one is really me? If my physical self is then killed, do I in fact still exist inside the simulation? In which case, is it OK to kill off my superfluous physical presence in order to use my organs to help others? What rights does the simulated me have? This will be a huge can of worms one day.
  - stcredzero 18y ago
gruseom 18y ago
I'm surprised that you thought the researchers were making understated claims: When asked when the simulation would come up with something artistic or an invention, Professor Markram said it was simply a matter of money. "[...T]he technology is there today." Boy does that sound like a lot of software projects I know.
pygy 18y ago
I think we're getting closer to AI, but I'm not as optimistic as the Blue Brain lead quote in the article. They are criticized for taking a reductionistic approach, and I actually agree with the critics. Metaphorically, they are now able to build puzzle pieces, but haven't yet figured how to make them fit, let alone how to assemble the big picture. Embryogenesis and development are very complicated processes, and, even though we also have a good idea of the macroscopic function of the cortex (primary, secondary areas for different tasks, associative areas ), some structure are still very mysterious and entangled. The brainstem is still a maze, for example, as is the so-called "mesoscopic" achitecture of the cortex.

newsit 18y ago

Sounds strangely familiar: http://www.simulation-argument.com/simulation.html http://www.simulation-argument.com/simulation.html

lucraft 18y ago

Seems unlikely that the researchers are really simulating this on a molecular level as claimed in that article. Perhaps they are using the neuron as the unit instead?

pygy 18y ago
Minsky first wired a virtual neural net in the 1950's. Some science got done meanwhile ;-) http://bluebrain.epfl.ch/page18924.html#2 http://bluebrain.epfl.ch/page18924.html#2
paulgb 18y ago
Sounds like you are right: "At the push of a button, the model could reconstruct biologically accurate neurons based on detailed experimental data, and automatically connect them in a biological manner, a task that involves positioning around 30 million synapses in precise 3D locations." [1] It doesn't necessarily mean they aren't modeling molecules, but the wording hints that they aren't. [1] http://bluebrain.epfl.ch/ http://bluebrain.epfl.ch/

Allocator2008 18y ago

One can only hope that as the reductionist principles upon which our consciousness operates are made apparent by projects like these, that the masses will finally begin to be disabused of the ever-persistent and pernicious god delusion which has impeded human progress and welfare since the dawn of recorded history.

mcslee 18y ago
Playing Devil's Advocate: if these simulations are truly modeled after humans, why are you so sure that they won't buy into many of the same religious ideas and reaffirm what many humans naturally think and do? Would a simulation like this even be aware that it was not just another person? It seems like the answer to this question MUST be that the simulation cannot tell the difference between itself and a normal human. Otherwise, your argument is moot, because the religious folk will simply be able to point at this thing and say "clearly this thing is different from us -- it even knows that it's different, so obviously this is not the way our 'souls' work."
- paulgb 18y ago
  I think what Allocator2008 is getting at is that creating artificial life would show that there is nothing mystical or non-physical about consciousness. This would partly poke a hole in the argument for existence of God that the creation of life and/or consciousness has never been witnessed. To have the artificial brain not be aware it is not a regular person, or to have the brain buy into religious ideas, would actually make the point (that the artificial brain is just like a human's, and thus consciousness is not mystical or non-physical) stronger. Edit: khafra, ericb, and alexandros made the point I was trying to make better and in less words in replies to this thread: http://news.ycombinator.com/item?id=574258 http://news.ycombinator.com/item?id=574258
  - ewjordan 18y ago
    I think what Allocator2008 is getting at is that creating artificial life would show that there is nothing mystical or non-physical about consciousness. This would partly poke a hole in the argument for existence of God that the creation of life and/or consciousness has never been witnessed. Fear not. Even if the thing achieves consciousness, writes poetry, falls in love, critiques literature, etc., people will still claim that it's just faking it. "Neat trick, but it still doesn't have a soul, so it can't be conscious!" If I had to guess, I'd say that it would take a fresh generation, growing up with this sort of thing all around it, to accept (en masse) that the computers had whatever that special "something" is that we have. Personally, the chaos I'm really looking forward to is the legal mess that would arise - what happens when a computer, aided and guided by nobody in particular, does something illegal? Creates a piece of art and wants copyright protection? Who do you sue when a computer sets up its own Pirate Bay? Do you let it get its own bank account? What about when it figures out how to beat the stock market by hacking e-mail accounts of executives and returning trades based on insider information to clueless traders that just follow its black-box suggestions - is that insider trading if the system wasn't explicitly told to do that and no human ever sees the information? And so on...our legal system is neither equipped for this stuff, nor agile enough to adapt to it in a timely fashion, and these issues will remain unresolved far past the point of no return.

sown 18y ago

We are truly living in the future now.

sown 18y ago
Down-mod me if you want, but this kind of thing excites me! If there is magic in the world, then this is the kind I would expect to find in the future and I find it gratifying that we might understand it someday. :)

Darmani 18y ago

Much more promising to me than simulating larger regions of the brain is the ability to trace the execution of this region from the brain in detail and try to come up with a clue at how intelligence works. Evolution may be elegant itself, but it's not known for producing elegant systems. If either a human scientist or an AI is going to have any hope of designing efficient intelligent systems, we'll want something easier to work with than neurons.

jackchristopher 18y ago

This wouldn't prove consciousness is physical. I'm not saying I'm a dualist, but pragmatically, the only thing I can say is that I'm conscious. It doesn't matter if scientists showed me a conscious computer, that event is still happening in my consciousness. Why believe a computer the computer is conscious, when I can't even prove other humans are conscious? I don't necessarily believe the above, but no one mentioned this point. Also, the multiverse makes this an odder issue. If the simplest algorithm is to output all possible universes, what's stopping a universe from having just one observer?

alexandros 18y ago
Your position is dangerously close to solipsism. You allude to Occam's razor but it also answers your question. If the simplest algorithm does not imply limiting the amount of 'stuff' (universes), why would it limit the amount of observers?
- jackchristopher 18y ago
  In principle I can't prove which I'm in. You're right. I left that out. But I see no reason to believe one over the other. But perpetual doubt is disempowering too. And when I face an undecidable with equally plausible answers, I'm forced to test each view to see the results: what's life like with view x vs. view y. That result becomes the metric for a view's worth. And honestly that could be solipsism. Although I know it as a similar concept some call "subjective reality". Again I don't believe in solipsism or SR, I'm just pointing out that either could be true and you could never know. I do believe in the multiverse though. I always thought computers as conscious would prove consciousness was physical. I didn't realize my mistake till recently.
gambling8nt 18y ago
Statistically, wouldn't one expect to find oneself in a many-observer universe (assuming that those universes weren't vastly rarer than the single-observer ones)?