The Brain in the Machine: IBM's Compass
newyorker.com
newyorker.com
We don't know what the information transfer mechanisms are in the brain, nor do we know how information is stored. We do know that action potentials are the observable and recordable signs of potential brain activity, but that's all we can see. If you have something as complicated [1] as the brain, assuming that one of the first things we found about the brain is the way it lets me think of art, reason about the past, and adapt to my surroundings is plain stupid.
Since we discovered neurons [2], there have been monumental changes in how we understand the world. But this still isn't enough. We cannot explain the things that happen in brains with our current understanding of physics. Minds are more connected than one may initially assume.
I'll tell you a story about why I know this to be true, and what it means for our attempts to recreate a brain.
For as long as I can remember, I have been falling asleep with some sort of music or movie playing.
Whenever I would drift off to sleep, I would hear this noise in the background. It let me sleep easily, giving me something to focus on. When I went to Norway this past summer, I didn't have anything to listen to, only my mind.
I noticed when I was about to fall asleep, something weird would happen. I would hear things that weren't thoughts I had, they were thoughts in other languages. They sounded like what I was hearing in Norway, but they could have been Chinese for all I know, but what I do know is that they were sounds I never recognized, and they were in languages that weren't my own.
Whose thoughts were these, and why were they in my brain when I was falling asleep? Weird things happen when we sleep, but before all contact from the real world shuts off, the transition into sleep may be one of the most important for figuring out what is really happening in our brains.
How would this information transfer boundaries between my brain and another person's brain, or another region in space time? This would be easy to pass off as misinterpreted noise, but this happened for many times while in Norway, but when I got back home, it tapered off (I lost my peace of mind when I returned home).
Our current understanding of physics is not advanced enough to allow us to understand the brain. We have done cool things with technology, but I think we are farther from understanding what is a person is in the true sense of the question. It's interesting to think about the 'hard problems' that persist over life times. Almost all the problems that we are solving today are not 'hard problems'. They are challenging, fulfilling, rewarding, and helpful in the short term.
I once went for a long bike ride, and it had been a while so I was pretty excited. I told my dad about my achievement. I said "Dad, I went for a 50 mile bike ride today!". He said "Yeah, but did you get anywhere?".
We are solving problems, but the solutions are not advancing the human race. We are riding around in a circle.
I'll stop now, and hopefully start a blog so I can put out my long winded rants there.
[1] http://www.cerebromente.org.br/n17/history/neurons2_i.htm
Why do you immediately assume they aren't your own thoughts? Surely you have no problem simulating garbled language or your own garbled interpretation of other languages in your mind right now, why couldn't this have been the case when you were falling asleep? It seems strange to me to think that because of some bizarre or out of character thoughts that the only answer would be someone else's thoughts invading your mind.
I don't think someone was invading my mind, I think I was detuning my brain and tuning into something else, a 'sea' of thought. I think they aren't my own because I can't consciously think those thoughts, I cannot simulate what happened just by thinking about it.
I think we have more control over the world than we give ourselves credit for. It's easy to cast off stories that are not backed by fact, but what if this was the divide? What if we just needed a little bit of faith to see something that most people can't see? What does a kid see when they play in a box? They see whatever they want to see, they are not tied down by what society deems as real, they create reality for themselves.
For a while, I had a spurt of thinking about some new tech idea and seeing it within a week of thinking about it. It was almost like I was forging the path through the sea of unknown information around me. I don't have enough experience in the world to determine whether or not my thinking had influence on what happened, but it provides an interesting framework for some thought. What if you could control the world, and you were the center of the 'relative universe'? Quantum mechanics predicts that prior to the first consciousness observed a wave function collapse, the world could have not existed, it was merely a fuzzy combination of wave functions that never had definition. What if this was happening every time I did something, interacted with something, noticed something. What if, when I thought about something, it collapsed the wave functions necessary to bring me that information? What if I controlled the world, and whenever I reached in to grab information, the necessary wave functions collapsed (regardless of where they had to). What if we could have multiple layers of collapsing wave functions, so you can live your life, I can live my life, and until we communicate, we lose the necessity of having the same time or space clocks? What if the world reconciles the differences when necessary, rather than all the time?
It's easy to get sidetracked when I'm thinking.
At what point do I call a thought my own? I'm not sure, I'm not even sure if ideas are my own thoughts. My thoughts are the ones where I can follow the train of logic that got me to where I am now. An idea is a lightning bolt. To backtrack where an idea came from, generally all I get is a couple inputs, a big jump in logic, and a conclusion. you can say what inputs were necessary, but that doesn't always get you to the output. I call these 'prime ideas', these are the good ones that cannot be assembled from combining (in some way, shape, or form) previous ideas. Prime ideas are ones that don't feel like they come from me, I wasn't thinking about anything, I was cleaning the kitchen when a lightning bolt in idea space struck my brain. Those don't feel like something I made, they feel like something that was given to me, like someone saying "Here, look at this picture" and I see something cool in my brain.
>The thing that made these 'thoughts' different is that they felt real. I experienced hearing something in another language,
- So do schizophrenics. That doesn't make their experiences any more concrete.
>What if we just needed a little bit of faith to see something that most people can't see? What does a kid see when they play in a box?
- Faith in things is important - probably intrinsic to our existence. It gives us things to aim for, to believe in. Systems of rationalisation. But believing something that is not possible will not make it so.
Kids are playing. Daydreaming. creating possibilities. You can do that when you're grown up too. I daydream all the time. But that reality doesn't exist outside of the story, the fantasy in the head of the creator. It could be argued that the greatest inventors and innovators (not to mention story tellers and artists) are able to mobilise their vision and dream into reality through faith. But there are still lots of things that can't, won't, exist.
>For a while, I had a spurt of thinking about some new tech idea and seeing it within a week of thinking about it
- This is a relatively well known phenomenon. - http://en.wikipedia.org/wiki/Multiple_discovery - It has happened throughout history. I have personally felt this a number of times - that I have come up with an idea independently - and then you see the idea out in the wild. On closer reflection I realise that I was more likely influenced by my extensive readings and my day-dreaming to make links between existing technologies, discoveries and inventions; formulate and write up an idea - but these ideas are in the 'potential space' of possible ideas and concepts; and someone was more on the cutting edge than me; and did it better and faster.
>What if you could control the world, and you were the center of the 'relative universe'?
- interesting, and i've had similar thoughts. Forever unprovable however; and the idea of us each controlling a 'relative universe' I think on some level is true - i.e. we can do amazing things if we just set our minds to them and rise above apathy and entropy; but on another; we still exist as part of the universe, which has universal rules.
we are the universe experiencing itself; there is so much we don't know; keeping an open mind is important; so is dismissing things that are not the truth; the hard part is filtering the wheat from the chaff.
Yes, the author is in error, because he has postulated an arbitrary solution (i.e., one without evidence) to the problem of hearing voices. And, yes, religion also postulates arbirary solutions.
tl;dr The problem isn't that he is going on personal observation and experience; it's that he's made an error in doing so.
You might be interested in watching a Google tech talk by Dean Randin (also wrote a book): http://www.youtube.com/watch?v=qw_O9Qiwqew
He seems to have been pretty rigorous when investigating possible subconscious psychic activity in regular humans. And it does seem like most people dismiss such things out of hand, which is the very antithesis of scientific thought.
Tangentially, I don't think we can call any brain state "well understood", never mind hypnagogia.
Seems to me that between awake and dreaming, as you come to accept more and more things as real then all bets are off. You could imagine hearing chinese and totally accept it. Sort of like a drunken person being convinced their judgment is just fine.
When it comes to consciousness the problem is not that we haven't built a big enough simulation with enough neurons, but rather that we lack an explanation for consciousness and its associated complex phenomena, like qualia. When we can explain consciousness, he suspects it will be easy to program. Conversely, if you can't program it, you don't understand it.
Tracking down such a thing sounds like trying to find the essential "Tuxness" of a given Linux installation.
If I install one kernel patch, does that make the system "less Linux-y"? What if that patch was a bug fix by Torvalds himself? Would that have the opposite effect?
I know the answer to the questions above: they're nonsense questions, so answers don't even enter the equation.
We have identified specific modules of the brain that, among other things, let us distinguish faces from mere shapes, friends from strangers (even when it comes to the same individual), and language processing itself. Our own sense of self seems to reduce to a distinct, toggle-able neurological phenomenon.
In short, the very idea of qualia reduces to just various artifacts of our brain's multiple systems of abstraction.
Simulating a brain using the same number of neurons is like simulating a body using
the same number of atoms.
Which boils down to "make everything as simple as possible, but no simpler". Chomsky wants to understand what's happening. Norvig says it's too complicated.
I think we can usually carve off an aspect or layer of phenomena that - in isolation - is simple enough to understand. But I don't see any guarantee that this will necessarily be the case; models are possible that aren't decomposable into simpler parts (because they must interact with everything else, informationally). However (1) understanding is what we seek in Science; and (2) it's always been possible before - for whatever reason, the universe does seem to be decomposable hierarchically... at least, the bits that we've understood, so far...
Create the biggest MMO ever made, in a week?
If you push is a bit, you can also branch the human brain at a certain point in your life and merge back to master whenever you see fit, and you can experience more than you ever could in a single life.
EDIT: I realize space is a big deal, but I only wish some organization working with the funds that NASA has that can explore the human brain and consciousness. This could initially lead to a variety of psychological hacks coming out which could be used by marketers and ultimately a deeper understanding of intelligence. Like OP said, if we could simulate a brain in a computer, there's no reason we couldn't make it better. If we could make it better, and this is pushing it even further, provide it with the right "memories" as a starting point, you could have the equivalent of the greatest minds of humanity living simultaneously. Think about what we could achieve then ?
Will there be movements to give legal rights to our virtual creations based on the notion that they are conscious? Will this cause humans to lose control of things as the computer minds grow more powerful?
Will we have laws that force us to imbue emotions in them? So that they can only be created with empathy for human beings?
Will everyone obey these laws when creating computer minds? Will there be black markets of unempathic, cold-hearted superminds that blindly obey their self-serving owners?
Will we all be out of jobs? Will we even need or want money?
But it bewilders me how so many people see the "holy grail" in simulating whole human brains. Two things:
1. Human brains are extraordinarily power efficient - I don't think we can beat them anytime soon, as long as we are power constrained. I believe the difference in efficiency for tasks the human brain is good at (such as visual pattern recognition) is a significant number of orders of magnitude.
2. Computers are at their best when they complement us, when they solve tasks the human brain is NOT good at.
I remember reading somewhere that humans might well become a completely digital species if this ever occurs. Interesting stuff to say the least...
We also don't have any idea how the physical structure of the brain maps to its functional characteristics (pattern recognition, consciousness.) This work may help us explore that. It's a brain simulation, maybe, but it's definitely not a 'mind' simulation at all.
It is important to clarify that we have not built a biologically
realistic simulation of the complete human brain. Rather, we
have simulated a novel modular, scalable, non-von Neumann,
ultra-low power, cognitive computing architecture at the DARPA
SyNAPSE metric of 10^14 synapses that itself is inspired by the
number of synapses in the human brain.We live in the age of simulation in science. Computation has reached the point where, in many disparate systems such as weather models, particle physics, protein folding, and neural architecture, building complex simulations lets investigators freely play with hypothesis and compare the results to real-world data. Hypothesis are proved or disproved, science advances, and we know that much more about how such complex systems indeed work.
The author of this article derides this approach, saying that we just don't have enough information to do anything useful with the machine. I call bullshit.
One of the entire points of the machine is to allow experimenters perform experiments that are either impossible or morally unethical on biological creatures.
In a few years, suppose a research decides to change the algorithm about how simulated neurons exchange a neurotransmitter or an electrical pulse. The simulation then produces patterns very similar or identical to an FMRI or electrocardiogram of a living brain. Biologists can then use that as an unexpected avenue to exist living neurons to see if they react in the same way. Its an instance of simulation allowing experimentation that illuminates the real deal.
Ignoring the fact that whether this ever happens is up for debate, why would it be moral to torture a simulated animal as long as you're not torturing a "real" one?
> In a few years, suppose a research decides...
You vastly overestimate what research will produce in a few years.
Morality is a human concept that should serve human ends--individually and socially. As such, performing experiments on a dog to gain more knowledge is moral.
This is a completely unsubstantiated assumption.
As such, performing experiments on a dog to gain more knowledge is moral.
This relies on the unjustified assumption that causing suffering of a non-human is not immoral or amoral.
This relies on the unjustified assumption that causing suffering of a non-human is not immoral or amoral.
That's like saying "God must exist, until you prove that he doesn't." In other words, on that particular point, you're the one making a positive claim.
Edit: actually, you claimed without justification to know that the action was moral.
In general, I am highly suspect of anyone who claims to know with certainty what morality is or isn't, or why it exists (or doesn't). People have been contemplating this for hundreds of years. It is very unlikely that you (or anyone) have discovered a new, empirically verifiable truth about morality. It is more likely that such claims are overconfident and unjustified.
So, rather than saying, "Hey, you're the one making a positive claim," I should have just said, "Actually, here is what is right, and here's why."
Of course, the "why" would be limited in scope, because I'd rather direct you to external information that I agree with than write a philosophical treatise in the comments of HN.
Anyway, I guess at this point I'm talking about how to talk, which is starting to get "too meta." Forgive me. To move on to something more interesting:
In general, I am highly suspect of anyone who claims to know with certainty what morality is or isn't, or why it exists (or doesn't). People have been contemplating this for hundreds of years. It is very unlikely that you (or anyone) have discovered a new, empirically verifiable truth about morality. It is more likely that such claims are overconfident and unjustified.
I basically agree with your analysis, here. But, unlikely things actually do happen. In this case, I believe they have. I have some info on my profile that will direct you to more information if you are interested.
don't presume either way until you have evidence.
It seems that if you were following this guideline, you would not simply state,
As such, performing experiments on a dog to gain more knowledge is moral.
Moreover, I am not sure how you are construing my claim as being positive. You said X is moral, and I said that such a statement depends on a source of knowledge that X is not immoral (that i.e., X is moral, your statement). There is no assertion I am making in there, except by implication, I am saying that facts about morality are like other facts, and need to exist and be demonstrable. You should be able to show them in some fashion. When you say dog experimentation is moral, you need to show the source of your facts.
Again I guess I have to rely on your claim to know what and why morality is, because you don't want to explain that.
Moreover, given that you seem to have some familiarity with philosophy, I don't understand why you would make such a comment about whether or not one ought to presume this or that without discussing the precautionary principle, which applies directly to the question at hand. The logic of the principle would seem to pose a direct challenge to your idea that one ought not presume either way.
I'm sure you've read some books or some arguments which you are relying on here, but I can't take your claim that human morality is X, and that Y is moral, on face value. And I cannot think of anything you could possibly be relying on as your hidden source for these statements which is not also dependent on a set of assumptions.
If life arrives on Earth from elsewhere, and is much more intelligent than us, let us hope they don't consider us to be dogs.
I think the same arguments that establish that humans have rights would still hold, even for the aliens. I also think that they would apply to the aliens, so that we should recognize that they have rights. But all that is just "possibly," because it's hard to speculate about a case like that.
A further note to that effect is the recent movement to give sentience-rights to cetaceans and the protection that goes with it. Its encouraging to see that movement taking off, as it will bode well for our hypothetical simulated fellows in the future.
To your second point, I believe you vastly under estimate the research that will be done in the next few years. There are incredible things already done in labs around the world that aren't widely viewed yet. In the next few years that research is going to start exiting the lab and start hitting the marketplace. The last 20 years of advancement are nothing compared to the next 20. We would all do well to bet on a stranger, faster future rather than one that seems like the status quo of the moment.
Odd. As my research is in simulation using massively parallel supercomputers, I'd suggest that unless there are (secretive) private research groups out there that are way ahead of universities and have access to pflops level computing, I don't think we'll be seeing accurate simulated FMRIs any time soon.
Weather simulations are generally reasonable, though limited in this regard. Other scientific simulations tend to be a little more hazy, the main aim of course being to produce a publication.
You said "Hypothesis are proved or disproved, science advances, and we know that much more about how such complex systems indeed work". Could you provide some examples? In my experience I've rarely seen simulations have have proved or disproved hypothesis. This is an honest question, if you have some references to significant results from simulation I'd be interested in reading them.
A popular debate about what knowing the brain's wiring diagram will bring us
"In sheer scale, it’s far more ambitious than anything previously attempted, and it actually has almost ten times as many neurons as a human brain."
No, it has ten times as many simulated neurons. It does not have any neurons. Unless it's a big wet biological computer.
This kind of sloppy thinking is not too surprising to me, given that this is The New Yorker, which I consider to be exemplary of sloppy thinking almost as a rule.