Simulated brain closer to thought | Hacker News Reader

Simulated brain closer to thought

news.bbc.co.uk

62 pointsalexandros54 comments

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.

And very interesting science.

"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?

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.

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.

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).

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.

Okay, these guys aren't anywhere near simulating an entire brain molecule by molecule. There's a big difference between this and being able to simulate the storage of a memory.

Also, the estimates of the total complexity of the brain may have all been greatly underestimated. We now have strong evidence from evolutionary Proteomics that each synapse has considerable complexity and is itself involved in computation. This makes perfect sense: the synapse must have evolved from the molecular sensing and navigation mechanisms of single celled Eukaryotes, and these are clearly capable of some computation.

(See the Brain Science Podcast episode 51)

I'm just extrapolating. Given exponential increases in computational power, it's a pretty small jump, I think, to imagine full brain simulation from what they're doing.

I guess that would depend on where (or from who) your morals come from.

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.

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.

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.

I really doubt digital computation will ever be a platform for intelligence as we understand it. Biological systems really are different than digital ones.

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.

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?

Minsky first wired a virtual neural net in the 1950's. Some science got done meanwhile ;-)

http://bluebrain.epfl.ch/page18924.html#2

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/

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?

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?

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.

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)?

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.

We are truly living in the future now.

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. :)

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.

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."

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

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.

> 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!"

Good point. Academics seem to have accepted that consciousness isn't magic for a while now, so building an artificial brain might not be so convincing for the rest of the world.

> Personally, the chaos I'm really looking forward to is the legal mess that would arise...

Yeah. I wouldn't even know where to start with figuring out some of those questions. I'm sure someone has written about this already, so if anyone knows who and where this has been written about I'd be interested in reading about it.

So let's see, you set out to build a human brain and emulate consciousness, yet in the process you reaffirm to everyone that it takes a phenomenal level of supercomputing power, complexity and intelligent, purpose-guided scientists to create something that can operate on its own.

I may be mistaken, but tell me again how intelligently designing and creating an autonomous entity furthers the atheist world-view (which I am guessing you agree with, given your careful choice of phrasing ala god delusion)? If I was one of the masses -- who are evidently far less enlightened than yourself -- I would argue that this project demonstrates that an intelligent being is clearly involved in the production of a conscious entity.

This is not about ID vs. evolution. It is about whether the brain can be reduced to a purely physical system or if there is something nun-physical about it, i.e. a 'soul'. Clearly, building a functioning artificial brain by copying the physical characteristics of a human brain should settle this one. I'm not saying we are there now, but there is certainly progress.

I did not say it was about ID vs. Evolution, either, but the original post clearly indicates a world-view that claims a lack of a designer. As I already noted, there would be an obvious logical dilemma between how the results of this experiment were achieved (if it ever came to fruition), and his world-view.

Clearly, building a functioning artificial brain by copying the physical characteristics of a human brain should settle this one.

You state this as if replicating a human brain would be like buying a new Lego set and following the instructions. I think we should first nail down the theory on why we sleep. You are aware that scientists still can't fully explain the most common mental process that takes up 30% of our lifetime?

The human brain will never be perfectly replicated in this lifetime or any lifetime, in the same way that a group of toddlers would never 100% figure out the full technology and complete mechanical blueprints if they were handed a Ferrari.

They think sleep is garbage collection: http://news.ycombinator.com/item?id=547598

Regardless of the question of sleep, the lack of a human known answer does not mean something else must exist that knows the answer. Perhaps the answer is unknown in the universe.

If you believe otherwise, where did the intelligent being who knows this information come from?

lack of a human known answer does not mean something else must exist that knows the answer

However, in your same vein of logic, the incapability to prove the existence of a designer does not invalidate the existence of a designer in the equation.

This same flimsy logic is applied within scientific naturalism, where we assume that because the existence of a designer is not capable of being proven, that the only way we can conclude something "coming about" is by inferring from only what is empirically given.

Surely you would not try to argue with me that because a child is not present (and thus logically unprovable to exist) that the existence of a sand castle on the beach could only be explained by the waves and functions of the ocean.

Obviously in that example we have external information to invalidate our faulty logic (i.e. seeing prior children build sand-castles), but it serves to highlight the absurdity that comes about when trying to infer from what is only empirically given. Now take that simple sand castle example and apply it to the insane complexity of trying to explain the creation of the universe. Not to mention remove the external information that enables us to validate/invalidate our claims.

I hope you can see why I choose to merge my intuition for recognizing intelligent design with empirical science.

There are many people on HN capable of making pattern-recognizing neural nets. Neural nets are loosely (very) modeled on human neurons.

To create a neural net, a framework is created that sets up layers of neurons and sets their weightings (often randomly). The framework runs the net against batches of data, measures the level of success, and changes the weightings. Eventually this process produces a "brain" that can, for example, recognize letters.

The framework's creator can't explain how the generated neural net functions because no one designed it. To someone not familiar with the process, the neural net would trigger their intuition for recognizing intelligent design.

OK, and so what does that conclusion have to do anything?

You just went in a big circle to tell me that my intuition about the neural network was correct -- that there was a designer (the framework's creator) outside of the neural network who set the boundaries and then let that unique neural network run its course.

It doesn't matter if we know how the neural network works (and so forth), because this argument becomes logically invalid the moment you inject a designer (The framework's creator) into an argument to prove the lack of one.

Edit: Even if your comment was trying to show how an intuition for recognizing design can be "fooled", it still fails out because the intuition was correct.

There was an intelligent designer for the framework, sure, but not for the intelligence in the neural net. That was just the product of a feedback loop organized by simple rules. Mistaking order (like a fractal, for instance) for intelligent design is one item I was addressing.

>> I think we should first nail down the theory on why we sleep. You are aware that scientists still can't fully explain the most common mental process that takes up 30% of our lifetime?

>> The human brain will never be perfectly replicated in this lifetime or any lifetime

The other item I was addressing with the example is your implication that we can't create something until we understand every aspect of it.

And finally, adding an intelligent designer can't do more than push the argument back a step and leave you saying "who created him?"

You are aware that scientists still can't fully explain the most common mental process that takes up 30% of our lifetime?

I don't think that's clear at all. I might agree with your first point, that we don't understand the internal mechanisms very well, but in some cases it's a lot easier to model a system than to understand how it functions. This is particularly true of emergent behavior, which pretty much anyone in the field acknowledges the brain is largely driven by.

Even in fields as concrete as physics, there are plenty of emergent results which can be verified to arise from the small-scale behaviors that we can observe, whereas a full "understanding" of how that emergence happens is beyond our grasp.

I'm not sure that it will settle anything. Even if you built a brain that functioned identically to a human brain, one can never prove that the brain is experiencing/creating consciousness. It could always be argued that the brain is just mimicking function but lacks subjective awareness. Since we have no way of measuring consciousness even in a human brain, how will be prove that a man-made brain is conscious.

Probably won't change many minds, but it does prove there's no "magic" in the recipe, and proves that god(s) are not necessary for the creation of life. It challenges the assumption that some magic chunk of soul-cloth is required to make consciousness work.

That said, surely there will be those that argue that conscious machines have no rights because they are missing their chunk of soul-cloth.

Philosophical dualism (http://en.wikipedia.org/wiki/Dualism_(philosophy_of_mind)) is a mistake at the core of most religions. Making a human brain would contribute much to destroying the concept of the mind as something separate from the brain, rather than the mind as something the brain does.

I would not call it a mistake, yet. Just another philosophical position (which I do not agree with, but that's not the point).

Apropos dualism: Did you thought about how artificial intelligence might end the dualism in economic theory: I.e. that one has to account for labour and capital as completely separate factors of production? (I am ignoring land etc here for the sake of simplicity.)

Interestingly this might just be a result of the perceived dual consciousness of the brain - see http://neuropolitics.org/Conservative-Left-Brain-Liberal-Rig... and others

dualism is also the core of the age of enlightenment and scientific philosophy. funny that.

I've never understood this argument. If an INTELLIGENT being produced the conscious entity, who or what was responsible for creating this intelligent being?

And no, you can't side step this question. You can't say "well, it's god" because that's special pleading. I just don't understand how people can assume that evidence of complexity requires a designer, but never seem to have a problem inventing said designer out of thin-air as a magically preexisting entity.

What's more likely: that a natural process of change followed by selection essentially navigated through an infinite search space of possible configurations, some of which would create things like intelligent life, or that there was, magically, a pre-existent eternal intelligent being that decided to create other intelligent beings?

The people who propose a designer simply haven't learned how to think properly. It's as simply as that. They don't understand logic even though they pretend to.

I imagine you'll find it pretty frustrating when people start reprogramming their robotic servants to believe in their preferred god.

That's what electric monks are for.

Do android priests dream of electric choir boys?

Douglas Adams never hinted at any sexual preferences of electric monks in Dirk Gently's Holistic Detective Agency.

http://www.mat.upm.es/~jcm/adams.html

It won't seem much different to me than programming your children to believe in their preferred god.

Precisely what makes you think that a group of people who dismiss a concept as simple and transparent as natural selection will accept a collection of--probably significantly more complex--"reductionist principles upon which our consciousness operates"?

Similarly (and in opposition to many of the opinions in the other responses), if faced with artificial consciousnesses, why do you expect them not to ascribe to them the same "soul" substance that they allow themselves (ultimately, anyway--early resistance is likely to be universal)? After all, it is immediately evident that people can perform an action that creates an entity with an independent consciousness; if humanity develops another such creation method, how would that contradict their theology?

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

Reply on news.ycombinator.com