A Neuroscientist’s Theory of How Networks Become Conscious (2013)

The examples you gave are mostly abstract and not physical. If consciousness is a physical phenomenon, these algorithms must be manifest before they can qualify. As for the game of life, the pieces on a physical board aren't actually interacting with each other with feedback loops. In a computer, the circuits are mostly feed forward and the algorithm is simulated rather than physically realized.

"Perhaps the most unsettling implication of this train of thought is that anything can be interpreted as possessing any abstract property, including consciousness and intelligence."

[1] http://www.frc.ri.cmu.edu/~hpm/project.archive/general.artic...

There's another perspective here. We always argue from a very anthropocentric point of view and measure experience based on actions and reactions that a human would perform. Ie. asking questions, seeing something and saying something based on what we see, etc. The supercomputer you described can hardly be considered conscious with regards to such actions. However it might be very conscious with regards to a different set of actions, ie. turning off some switches which allow for intercommunication between the different CPUs (the computer might react by rerouting some packets), it might react to raising the room temperature (by spinning up the fans), it might react to removing some RAM (by allocating more data on the hard disk), etc.

If you implement a software system that is conscious, but then suspend it to a storage device, and resume it one week later (so that it is surprised: where did the time go?) its complexity hasn't gone away. It just wasn't conscious for one week because it wasn't running.

Another thing to consider is this: there may be more than one consciousness in your brain! The one which is you, the one typing on Hackernews, is just the one which has access to the "console" so to speak: the fingers, the eyes, ... There could be other consciousnesses hidden in your brain that are suppressed. Like "background daemons". Maybe that part of the brain that regulates your body while you sleep (or don't sleep) is also conscious!

The consciousness-process which is "you" is deactivated during sleep, but the other background "daemons" remain conscious. (Clearly, sleep is not a complete, global shutdown of brain activity, in other words.)

(Also, I'm here reminded of dolphins which put half their brain to sleep at a time.)

1. Your nearest definition of consciousness "emergent properties of lower level functions" or something along those lines doesn't really cut it, either. To say that something is an emergent property of another thing, without showing the steps of that emergence is kind of like defining red as "an emergent property of the sun and a sheet of cellophane". It's HOW it emerges that's the hard question.

2. If you were pleased to hear another person's discovery that "my experience of experiencing is irrefutable", you might like to read Descartes's account from over 400 years ago in his Discourse on Method.

He veered off into some strange and tenuous territory, but the opening of that book is excellent and it uses that discovery (a.k.a. the Cogito - I think, therefore I am) as the first axiom from which to build a whole philosophical system, and it's a really interesting read.

This is the guy who invented algebraic geometry, from which we get the Cartesian plane, and the idea of ordered pairs, which is such a fundamentally useful idea that it's hard to see just how brilliant a discovery it was. He had some things to say about consciousness that have not yet been improved upon.

EDITED to say that the book I was talking about was Descartes's Meditations on First Philosophy, for anyone interested. He decides, "let's doubt that anything at all exists, just doubt everything - where does that take us?"

Intelligence is the capacity to acquire and apply knowledge. Knowledge is acquired when it is represented in a system. Action is the application of knowledge. Thus, all systems in which actions occur are intelligent.

The potential for intelligent behavior is determined by the existence of a goal, the opportunities to acquire and apply knowledge that are relevant to the goal, the resources available to acquire and apply knowledge, and the capacity of the knowledge representation to express truths. With the exception of the goal, adjusting these constraints scales intelligence.

Intelligence requires a goal to determine action selection. No actions occur in systems without goals, and therefore such systems are not intelligent. The goal of a system does not influence the scale of intelligence of the system (which can be thought of as the magnitude of a vector), but only its purpose (which can be thought of as the direction of the same vector).

Opportunities to acquire knowledge range from stones falling on levers (the force exerted on the lever – from the perspective of the lever) to humans touching hot elements (nerve impulses from the sensory system – from the perspective of the human mind). No observation is fundamentally superior. Representations of observations (as knowledge) determine potential applications.

Opportunities to apply knowledge exist when it is possible for a system to perform an action. At the beginning of a game of chess, each player has many opportunities to apply knowledge. After the losing player has made his/her final move, the player has no further opportunities to apply his/her knowledge. Opportunities to acquire or apply knowledge which are not relevant to the goal do not influence the potential for intelligent behavior because they will not influence action and will not be applied, respectively.

Resources to acquire and apply knowledge make action possible. The resources available to apply the laws of nature are typically presumed infinite (mathematical models preserve truth and the laws of nature conserve energy, but just as mathematical models require resources for the completion of proofs, it can be assumed that the laws of nature may require resources). This differs from the limited resources available to computers and human minds.

Knowledge representations range from the experience of observations (the force exerted on the lever and the nerve impulse) to stored, abstract concepts (mathematical models written on paper and memories in a mind). All knowledge can be expressed as truths. Knowledge representations express truths with respect to the perspectives defined by their grammars. Knowledge representations with wider perspectives (fewer assumptions) are subject to fewer biases, and are therefore more expressive.

The degree to which an intelligent system observes itself determines its level of consciousness. Most humans are not capable of independently observing which of their neurons are firing at a given moment, or the location of the electrons traveling through their electronic devices. The availability of such observations would result in a higher state of consciousness. Perfect observation would allow a system to be completely aware (although attempting to observe all of a system at the same time, including its state of observing itself, is an infinite set), resulting in the highest possible state of consciousness. Intelligent systems that are conscious of their action selection are able to optimize this process, subject to the capacity of their knowledge representation to express truths.

On the deeper observation here: there are plenty of things that can exist, but whether they actually have being in this universe is a trickier question.

> It's really astonishing to me that he words it this way, because independently I have thought almost exactly the same thing for a long time.

This is just the famous "cogito ergo sum" of Descartes, dating back to the 17th century. A significant piece of epistemology is devoted to criticizing this reasoning:

http://plato.stanford.edu/entries/descartes-epistemology/#4

I find consciousness perhaps the hardest of the hard problems, in particular the self-awareness aspect of the phenomenon. I experience it, as I suspect (but can't prove) do the rest of HN's participants. But damned if I can figure out how it emerges from organized systems such as ourselves or even whether it's necessary.

I can imagine going about my day emitting behaviors without self-awareness: stimulus/response, variation/selection/inheritance over the short and long term yielding the rich behavior we associate with ourselves, our conspecifics, and the other critters we share the planet with. I can imagine everyone and everything else doing the same - behaving, absent awareness. So far, I haven't seen how consciousness is required for behavior.

And yet I feel aware, never mind that I don't see why that's required. Inevitably, that suggests to me a continuum of consciousness, from less organized to more organized systems. For me, it's a long walk from the Nernst equation and neuronal membrane potentials to my dog's happy disposition, which algorithmically seems to be wag more, bark less.

It's good there are hard problems to think about.

http://en.wikipedia.org/wiki/Orchestrated_objective_reductio...

I think it just goes to show how sick and deficient that worldview is. Have you witnessed that cliche in some films where a character confronts a scientist and says, "But there's more to life than cold hard facts! What about love and art and seeing your child go to school and ...!" And all the scientists watching the film scoff and think that's just fluffy and inexact and those are just emergent properties anyway. And it's true, it is fluffy and inexact. What the character should say is, "What about the fact that I experience my own existence!", just like you've said it. And then, afterwards comes love and free will and personhood and all the other stuff. But the hard problem of consciousness is the immaterial's undeniable "foot in the door". If you are interested in a true philosophy of all that is, pay attention!

Is also testable on kids: before the age of 3 they are unable to identify themselves as the people in any reflection. They slowly acquire this ability plus all the other traits we cluster in the word "consciousness"; including things such as conscious self-preservation, language, adaptive/abstract thinking, etc.

>If the idea bothers you that things might exist that are incapable of being verified consensually by society

This is nitpicking: But it depends way to heavily in what "society" means here, there is still people who believe heart is 6000 years old; so is never all people. the majority of people knows nothing about nuclear physics, and that doesn't make it false nor non-science.

What we have here is a system receiving inputs, observing how it reorganises itself under these inputs and influencing this reorganisation based on the results of previous experiences. For sure we can measure that a system doesn't merely learn but influences its own learning process depending on previous learning experiences. This may be what allows us to project knowledge from a domain to another, and what allows us to produce artificial knowledge by learning from others. We produce arbitrary connections through memory / continuity of experience and communication. This would provide a basis and a reason for language and societies.

What we miss here is a self-sense, and Koch hints at it IMHO. Integration may be key. I'd say dependency to make it clearer, state dependency. Two linked states where one reflects exactly the value or the other could effectively be labeled as one. They would feel as one, too. What I define as "me" is what has direct influence on me and what I have direct influence on. There isn't a clear boundary though. This definition still has a problem as we define me as what is related to me, so maybe a better definition could be found by taking it by the other end and defining as feeling like an individual "me" every intergrated conscious instance significantly disconnected from the others. A fragmented big abstract me, for short.

If we define consciousness this way it exists in the positivist framework (which satisfies me a lot) and yet it can exist in more exotic frameworks too, like the ones where one can experience that everything ultimately depends on everything when seen under the right scale, and that the sense of self can be extended to everything.

I'm still unable to devise a clear experiment to prove consciousness defined this way though, maybe because I'm plain wrong and it's impossible !

I believe Koch is conflating many distinct concepts - as most of the consciousness research field seems to.

How can you say anything useful until you define your terms?

It seems to me humans have many different layers of experience:

Visceral pain/pleasure Instinctive drives Emotional/limbic experiences Locational (physical mapping) Memory-based/learned Abstract/intellectual/symbolic

Consciousness is another kind of experience which adds a symbol that works a bit like JavaScript's 'this', and seems to create a level of experiential indirection - so instead of sensing something directly, you experience how it affects the 'this' symbol. It seems to do this by abstracting the experience and simultaneously comparing it with an associative database of previous experiences, using all of the circuits in the list. (And probably others too - I won't pretend this is a complete model.)

You need some basic ability to abstract to have a 'this.' The better you are at it, the more bandwidth passes through the experience/memory/abstraction/future prediction/new memories system.

Simpler animals have basic input circuitry, and they learn instinctively without abstraction.

But... I'd guess surprisingly many animals of all kinds have a 'this' network. The human advantages are a 'this' network with more bandwidth, a much improved ability to learn from experiential invariants, an ability to make useful predictions, and an ability to store and communicate 'this' experiences using external technologies.

If this sketch is accurate, consciousness doesn't have to be mysterious. But it also doesn't depend on a simple connection count.

The Internet won't be conscious because it has no 'this' network. It has no pain/pleasure circuits, no emotional circuits, and no instinctive drives. But most of all it has no ability to compare memories with current experiences and to abstract learning from that. There are systems - like Google search - which have incredibly basic precursors to abstraction/learning/prediction. But they still mostly run on a batch basis with input->processing->output, and no explicit self-concept state memory.

My guess is that if you built a system from these elements and included the current state of the art - or better - machine learning, machine vision, and natural language elements, with useful analogs of instinctive drives and the other animal fundamentals, you'd potentially have something that at least acted and sounded conscious, and would likely do some unexpected things. (Which is one of the practical ways we think we recognise consciousness. Conscious entities are hard to model. In fact we're probably easier to model than we think we are, but face to face, consciousness in other people usually implies surprises.)

Whether or not it would be truly self-aware or just pretending is a question that's impossible to answer without resorting to metaphysics.

I find it interesting to replace "apple" with "UFO", and you get a scenario that is no longer theoretical.

Right, but I don't think this is as unrecognized as you think. Entire systems of ethics are built on the assumption that since I'm experiencing existence, and you are roughly the same as me, you're probably experiencing existence too.

Some systems expand on this and stretch or restrict the definition of "same" from "skin color" to "has a central nervous system" and other notions.

The longer I'm around, the more I start to think that this experience can be really different. Quite a bit of research is going into some fragments of this topic, we call it "intelligence". And we come back surprised whenever we find out that something which we didn't think of as "the same as us", like an octopus, can have it's own "experience of existing".

More recently there's a growing body of work that's starting to show that things even less similar to us might have an experience in their own way. We're finding out plants, for example, have all kinds of interesting and subtle ways of responding to the universe. People are afraid to call this "intelligence", but there's definitely something there.

It may be more that "life", of any kind, can be considered under this notion of experience, and not just self-similar things.

A different spin, "you are the universe experiencing itself"

A problem is that when we use concepts to describe new concepts, lots of little assumptions tag along each referenced concept that may be different to every person. So, when I explain a concept, many people may understand my concept differently, while some may understand my concept precisely.

There is a difference between 'what could exist' versus 'what is assumed to exist'. When we communicate, we have to be very careful about the definitions of terms and their context.

I say "tree", we both may think of a tree in our minds. Your tree may come with assumption that the tree must have leaves in order to be a tree. My tree may come with the assumption that a tree does not need to have leaves, in order to be a tree. Neither of us may touch upon this point, while we are busy communicating about trees, and I think that is what is most important to recognize.

The rest of what you say though, did click for me, so thank you.

You can find common ground between the statement you quoted if you want to take solipsism as your definition of "consciousness." IE, experiencing your own existence is consciousness, or is the test for it.

That isn't necessarily a bad definition. I quite like it.

What I've never liked is this CS Philosophical debate. It devolves into semantics so often it makes me suspect that's most of what it is. We define consciousness in terms of what it does, in which ace we may discover multiple solutions. Or, we can define it in terms of how it does it, in which case we'll have to wait until it's discovered before we can see what it does. Either way, the word consciousness is not useful enough to merit all this trouble it causes.

Contemplating its own undeniable here-ness, that's the kind of spiritualism that separates bloodbags from greaseballs, for now.

Your "of course" is taking as proved something that is not (the consciousness is computation hypothesis).

In your benefit you said may so of course you're not convinced. Which is good because people shouldn't believe in things that aren't testable. :)

And you talk about birds but there are als unicellular organisms reacting to anestesics, so that doubt should be even broader

In fact, it "decides" for itself which ones should go viral.

Interesting thought.

BTW one doesn't even need to take acid to realize this. Humans have thought this for thousands of years. It's a common belief of ancient eastern philosophies (metaphysics) such as Taoism and Buddhism, that "we're one consciousness". Our egos just don't think about it, much like a cell of your body operates in another (lower) level of consciousness, where it thinks it's "independent of you".

I feel not entirely convinced that I'm a Harvard architecture. A von Neumann architecture seems more plausible... long story short because humans have huge problems learning to walk on two legs.

Evolutionary neuroscience might provide the answer we seek, through the normal technique of belief propagation.

http://en.wikipedia.org/wiki/Ship_of_Theseus

When you wake up after a dreamless sleep, are you the same person, the same conscious entity, as went to bed the night before? Or is that entity now "dead", and "you" are a new entity that has just inherited its memories (most of which it in turn inherited from its predecessors)? How could you ever tell? In fact, are you the same conscious entity from moment to moment, or at least from thought to thought?

More "making a copy" thought experiments, none terribly original:

- If you're disintegrated and immediately reassembled, are you still you?

- Does using different atoms make a difference?

- Does leaving a gap between disintegration and reassembly make a difference? If so, how long a gap? What if you're resurrected at the Omega Point by sufficiently advanced aliens/post-humans?

- If you're split in two (sagitally, coronally, or however), and each half is immediately reconstructed into a whole human, each identical to you before the split, which is you? Which pair of eyes would you find yourself looking out of? Both? Neither?

- If the two "yous" exchange atoms, such that one ends up with the entire complement of atoms that made up you before the split and the other ends up with none, does that affect the claims of either to be the "real" you?

If a program sent out a malformed packet, this would indicate an error in the sender's program. Let's assume that a receiving program reads the malformed packet and is able to process it. It results in something favorable (no clue how this would be determined, but let's go with it) and so the receiving program repeats whatever it originally did to get this malformed packet sent to it, so that it gets another malformed packet. Multiply this by thousands of other programs doing similar things to "hack" other programs into sending malformed packets to perform useful functions, and maybe you have something happening here.

The above process basically describes a genetic algorithm, which I don't think the Internet is. It also assumes that most programs are flexible enough to produce bugs like this - creating malformed packets. Does this mean we should stop unit testing code to allow more freedom from the programs? Plus, the above also assumes there is some way of ranking an outcome from a malformed packet as "favorable". Maybe that would be humans?

That being said, it is hard for me to imagine what a proof that charge is intrinsic to an electron would even look like. Proving that something is the result of something else seems much more straightforward than proving something not being the result of anything.

Also, only talking about an electron's charge being intrinsic is ignoring a lot of the mystery of it. The fact that it is a point particle is a pesky issue as well. An electron is just an electron with no underlying "pieces" or mechanisms. Obviously that could be wrong, but no one has been able to prove it yet.

I think using the electron as an example for our understanding of how the universe works is pretty ignorant. It's one of the weirdest parts of physics I've encountered during my career. It is a topic that few people actively work on but feels like their is clearly something amiss with our current understanding.

Moreover thunderstorm network (if we consider the interactions of particles in the thunderstorm) is very transient. No feedback loops emerge in the system (to my understanding of thunderstorm). Also the system does not adapt to the surrounding environment through it's reconfiguration of internal connections. Thunderstorm lacks many properties of the complex systems discussed in the article.

Also as theory is essentially useless as it does not predict anything.

They are throwing darts in the dark and making it sound cool.