Wednesday, April 23, 2008

Consciousness is Not Physical: The Mind Node

My last post contained the unavoidable fact that consciousness isn't physical. If I'm not making sense you can read about the problem again and again. Now I'm going to talk a bit about what it might be in the positive sense. But first, a little bit of the completely different.

I often read articles with novel ideas not mainly for the ideas, but for the techniques used to formulate them. I really like stealing new thinking techniques. As a courtesy for anyone like me...

The idea below has an unusual property. It supports my ideas a little too well - as if I had come up with it first, and then formulated the theory it's supposed to illustrate. It looks like a classic case of confirmation bias.

However, this is temporally impossible. In fact, I came up with the mind node immediately after I proved that consciousness isn't physical.

So quickly, in fact, that I don't remember how I came up with it. (Reviewing some of my notes, I realize I was thinking the first part up up a second time, to refine the concept. During this I must have suddenly realized how it would work.) I recognize certain pre-existing elements, but I was so deeply enmeshed in thought that I don't remember what I was thinking.

By contrast, when I finally proved that consciousness isn't physical, I was meditating on the exact difference between a machine and a mind - what feature, exactly, the machine lacked. You can see the how my approach organically led to the structure of the proof.

I've noticed that I'm not unique in this process. Many very good ideas, such as the mind node, come about in a state a mental 'flow' which seems to inhibit self-awareness to the point that memory formation is...not inhibited...more like irrelevant to the process. Memory becomes unimportant. You don't set out to do this on purpose. It just sort of...happens with you.

In this case, though again I don't know how, I realized that an earlier proposal that I'd read that tried to use quantum randomness to create free will could be modified, using the fact that information processing is represented arbitrarily, (a primer on encoding) to produce a system that is both causal and acausal at once. A system that expresses true randomness and infinity through finite components.

(Sorry, I didn't save the proposal I'm working from. I'd link it otherwise.)

I call it a mind node. I'm going to describe the overview and then going into all the details.

This is a diagram of an abstract mind node.

It has two basic components, the randomness generator, here depicted as a pentagon of five possible quantum states, and an interpreter.

Technically a mind also requires memory, though the interpreter ends up being a kind of memory.

The mind node functions in cycles. First, the pentagon is measured, leading to a random state. This information is fed into the interpreter, which interprets it. The interpretation changes the interpreter itself for the next cycle, and biases the five possible states of the pentagram.

The equation is an example of a particular possible start-point for the interpreter; x-q are the five possible states of the pentagon, C is a constant. The vector (x,y,z,p,q) describes the probability that the pentagon will collapse into each of the five states; f_r() is an abstract recursive function - the part of the equation which describes how the total equation changes each cycle.

Setting f_r() to zero, this example is essentially a rotator. If the pentagon collapses to x, then there's an 80%+C chance that the next state will be y. Because of the inherent randomness, this rotator will jump around approximately once a cycle, 20% of the time.

The math for this, assuming I haven't made a mistake, is exactly the math of a completely unpredictable system, it is the physical expression of true randomness. The state of even this simple rotator becomes progressively less certain with each cycle. The math is exactly the opposite of normal probability - the chance of the node being at any particular state at a particular time is determined by the product of all previous states, instead of being completely independent of the past.

If we then take f_r() to be not zero, the process compounds - the interpretation, and how the interpretation changes itself, depend on all the previous random actions of the pentagon.

Such a system can literally encode anything. It has a domain of five states and a range of infinity.

The infinite range is crucial, as we'll see.

A mind node attached to any computational network, such as by reading off the bias vector before it's sent to the pentagon, will effectively make the entire network completely unpredictable. Consciousness is infectious :-)

True randomness should be physically impossible. Consider a set up where an electron in a one-slit experiment has a target, a one metre square plate, which it can hit anywhere. Assume there's no restrictions on the probability of the electron - it's actually free.

At first, the measured average of the electron will move all over the place. Each new point will take the average in new directions. However, even if the electron decides (it's free, it can) it wants to pull the average in a circle, after a large number of impacts the average will settle down to be the center of the plate, requiring more and more off-centre impacts to move it. Despite the apparent freedom of the electron, we can predict it to some degree.

True randomness requires that the electron have an infinite canvas. In this case, even though the average of all impacts will exist at every point in time, there's nothing stopping the electron from moving it drastically. For example, let's say that the average is, in (x,y) terms, at (0,3). Let's say this is an average of 100 separate impacts. So, even if there were 100 more impacts at (0,9) the average would only move to (0,6). However, the electron can always hit, just once, at (0,1 009 100). The average is now (0,1 010 000/101) = (0, 10 000). This fact scales. No matter how many impacts there are, the electron can overwhelm all of them at any time. It is actually completely unpredictable.

Because this is a delicate point, I'm now going to say the same thing from another perspective.

Some people like to imagine that quantum randomness, called stochasticity, would allow free will to escape the trap of physicalist determinism. This is untrue. Quantum events are unpredictable, but not completely so. A large number of events will form an average. This is what stochastic means. This average is uniquely determined by the system. True randomness, by contrast, is completely unpredictable.

While yes, any particular event is unpredictable, they eventually have to flesh out an average. A very definite average that can be calculated in advance. Said another way, being enslaved to Newtonian determinism isn't worse than being enslaved to slightly unpredictable randomness.

So stochastic behavior has an average, a distribution, and all that jazz. All very concrete. But what if the average moved? What if the average itself was somewhat random?

We wouldn't even notice. The new level 2 randomness would have an average, which would simply combine with the first average to produce a single distribution with a single centre.

(And indeed in the case of a one-slit experiment, it wouldn't make any sense. This second average would somehow have to operate on a timescale different than the electron impacts. For example, electrons hitting every second would have to be affected by a process occurring every 1.5 seconds.)

What if we had an infinite series of averages, all of them moving? No single average would ever settle down. Each successive average makes the system more and more unpredictable - it widens the range. An infinite series of average is the limit of this behavior - it would have zero predictability.

Naturally, like all infinities, this shouldn't be physically possible. With anything that's not a mind node, it is impossible. The requirement of an infinite canvas is impossible to overcome.

I'm going to analyze the mind node again in the light of the series of averages.



Assume the mind node is initialized to be neutral - each state has a 20% chance of being measured. I'm setting f_r() as zero for now. Average level 1.

Run it once. Now, there's a 20% chance for each state to have an 80% chance of being measured.

Run it twice. There's a 20% chance for each state to have an 80% chance of having 80% chance of being measured.

Again, exactly the opposite of normal probability - each event depends completely on all previous events.

It looks exactly like what a completely true random event would look like. After each event, an average can be calculated - but it keeps changing. After an infinite number of runs, it would even reach the limit of having zero probability of being anything. Presumably, there is a quantum of probability. As such, only a finite number of runs would be required for the mind node to have less than a single probability quantum for each of the possible states.

Einstein taught us that two things that appear the same have to be the same. Things are defined by their interactions, which can be exactly described by math. The math for this is exactly the same as that for a spontaneous, true random event. Thus, it is a spontaneous, true random generator.

Combine this with a nonzero f_r() and the mind node can produce arbitrary sequences of bits. It can literally do any mathematically possible calculation, give any output.

Note that, while the math of a mind node exactly describes true randomness, at any moment in time it's completely causal. In short, a mind node breaks and upholds causality at the same time.

This is exactly what is required for interfacing with a separate physics. It must obey all the laws of the first physics, yet require outside influence to determine its behavior.

Further, a mind node would display all the behavior expected of a conscious entity. Moods, random deviations from expected patterns, favored behaviors (as the randomness settles on a particular region for a while) creativity as the interpreter explores new spaces, and so on.

Hopefully at this point, dear reader, you understand how and why a mind node works. Now I'll describe what it means. Also, as promised, I will discuss the consequences of nonphysical consciousness.

Discussion

First, as I mentioned before, since consciousness isn't physical there's no problem of Free Will. The brain is bypassing physical determinism. Nevertheless, it does not disprove Determinism. Rather, it means you can believe whatever makes you happy - there is no definitive evidence either way.
This discussion really annoys me, actually. The concepts aren't even physically valid and yet people insist on pouring on and on about the issue. [rant] It doesn't matter! There's no physical consequences! Not to justice, nor morals, nor feelings! It's a completely pointless discussion and I've proved not once but twice that you can't ever get anywhere in it![/rant]

Second, it means that evolution is even more impressive than we thought. Personifying it, it means that evolution takes advantage of not only every feature of physical law but actually transcends physical law, bringing something more into our universe.

Most impressively, it means I've solved the hard problem of consciousness. As a bonus, I've also solved the easy problem.

Fourth, it means that the difference between objectivity and subjectivity really is a profound qualitative difference. Literally, it means that objectivity is the realm of physics, while subjectivity is the realm of something I glibly call mentics, after the adjective mental.

Fifth, the proof does suddenly allow the idea of God, intention-manifestion, souls, reincarnation, ESP, etc... Nevertheless, it certainly does not provide any evidence for these ideas, and indeed a proper description of the subjective realm may simply re-prove that they're impossible. Indeed, since the subjective realm is almost certainly subject to self-consistency, we can rule out such objectionable things as the biblical God. Miracles are still impossible. Contradictions still don't exist. If they did, contact between the subjective realm and the objective realm could easily destroy the objective realm. Completely annihilate it, actually.

There's also some new, interesting questions. When you sleep, are you unconscious because you sleep or do you sleep because you're unconscious? The various answers have some profound consequences.

The mind node is equally fascinating.

I've already mentioned how even one of them would display moods and preferences.

It also handily explains why AI is so hard to do. A 'true' AI would be conscious, which is impossible without at least one mind node, and probably requires millions.

The state of any circuit computationally attached to a mind node would, as a whole, be a true random system. This has some fascinating possibilities.

For instance, it can explain why you can feel someone looking at you. Your mind nodes (you probably have something like a billion of them) and theirs become computationally linked. By seeing someone, their output is linked to your input, and you become single system, to a degree. Because it's all subjective, you can feel that click as it happens.

Similarly, it falls right in line with the fact that cutting the corpus callosum results in two people inhabiting a single skull. It's two sets of nodes that are no longer linked, which leads to two separate consciousnesses.

But, while I've established that a mind node is a hole in the law of causality, (which is identical to the fact that the universe is mathematically describable) the question remains, how does it restore causal closure?

Obviously, it somehow downloads information from the subjective realm. Similarly, it must somehow upload to the subjective material.

There's only one upload method possible in the mind node. It has to upload using the interpreter's feedback to the pentagon. This inevitably means that the particles themselves are conscious. It means that the electron in the pentagram can feel its probability distribution, and can further choose to collapse to any of them.

This is why consciousness isn't really a problem for physics. In reality, every particle in the universe is self-aware. Quantum randomness is consciousness, and it's both the upload and download method. It is the conduit through which consciousness affects energy and vice versa. Score one for Buddhism, I guess.

In fact, rather than being a problem, particle consciousness elegantly solves both the problem of decoherence and the problem of qualia at the same time.

Penultimately, the mind node illustrates why we need a physical brain to be conscious. The mind node does something very specific; it links the self-aware particle's output to a system of memory. Without that memory, the moment-to-moment existence of the conscious spark would be lost. Apparently, the subjective realm has no useful analogue to information storage and retrieval, and needs to employ physical structures. Similarly, freedom is impossible in the objective realm; it requires the influence of the subjective. In short, evolution has some strong reasons to evolve consciousness, and to ramp it up to the levels seen in human beings. Further, that consciousness depends so heavily on the state of the brain is not evidence for the physicality of consciousness. It's exactly what I would expect.

Ultimately, though, is the question of what an actual mind node looks like. I'm going to put that off and talk about the weather.

Climate is almost a mind node. It has an interpreter, the weather at present, and every single particle is a pentagon. The interpreter changes based on itself and on how the particles react to the forces of the present. However! The particles don't properly respond to the forces of the interpreter. Their state depends mostly on the ambient temperature. Similarly, with such a large number of interactions, the particles rarely reset to a superposition that can randomly decohere. Finally, there's no actual memory. The weather is incapable of recall.

Evolution is also almost a mind node. It has DNA strings, which randomly change, and are affected by previous states, and an interpreter (DNA polymerase plus physics) which nominally depends on the DNA itself. It even has memory in the form of the organisms which arise. However, the interpretation of the codons almost never changes. The thoughts of a consciousness using DNA as the pentagon would never change - the incoming information is constant. Also, the memory is not very robust. Storing memories would consist of the state of a species, and thus would corrupt very rapidly compared to the cycling frequency. (Our perception of a moment is almost certainly a result of our own mind nodes' cycling frequency.)

So near-mind nodes are prevalent in nature. They pop up often enough that it's somewhat strange that there aren't more actual mind nodes. Still, it's easy enough to build one in a lab.

We've been measuring quantum randomness for a while. Set any convenient system up, and then link it to a field programmable gate array (FPGA) which can program itself. Such a system will be a mind node and should display behavior reminiscent of consciousness, such as moods.

However, the initial condition of the FPGA interpreter are very important. There are lots of self-modifying equations that eventually settle out, especially since the number of equations that a normal FPGA can represent are limited by its logic gate count.

There's four classes of possible equations. First, equations that settle down to a constant equation. Second, periodic equations. Third, completely random equations. Fourth and most interesting, equations that display order but not consistent order. You can see these in cellular automata, and exactly the same kind of thing will occur in a mind node. Only class 4 initial interpreters will create useful consciousness.

Similarly, because of the random nature of the pentagon, even the most well-chosen equation will likely end up either chaotic or broken.

In fact, this is probably what happens if you stay up too late. As your sensory mind nodes accumulate garbage, the interpreter (which determines what you actually sense, remember) starts getting out of hand. In a brain, it will start co-opting more and more neurons, growing like a fungus. It's behavior will become more chaotic and self-referential rather than accepting outside input well. It will damage memory formation. In short, you stop being able to think and start being able to hallucinate.

So sleep is necessary probably because the mind nodes in your consciousness modules need to be reset every once in a while. So why do you dream? Test runs. Having reset the interpreter, it needs to be tested and tweaked a bit. With a poorly defined interpreter, the sensation it creates will be somewhat random. Once the diagnostic/reset is complete, you stop dreaming.

But what, exactly, is a mind node in the brain?

The brain is immune privileged, (which means it doesn't have the full range of features) just like the gonads and uterus are. They also share the feature of dealing with nonstandard DNA strings, which are usually attacked aggressively.

As I brought up before, DNA changes randomly when copied. What if the brain used low-fidelity copying to perform the pentagon function? We already suspect DNA fragments are being used as memory in the brain. This could be because they're actually past computations - run them through the proper interpreter and you think the thought again. The interpreters, of course, are neuronal pathways. Because of the massive parallelism in the brain, each neuron can be part of several interpreters at once, and because the brain's connections are completely plastic but depend on previous firing patterns, we have exactly a mind node. Incidentally, yes the mutation rate in different parts of DNA is different, and can presumably be tuned.

In short, having invented a mind node which breaks causality in exactly the way necessary from my proof that consciousness isn't physical, I found all the necessary components for the mind node in the brain.

Tuesday, April 22, 2008

Consciousness is not Physical: Primitive Version

What is consciousness?

Are you conscious? Did you evolve?

With the available evidence, we can now prove that consciousness isn't physical.

(Abstract: Consciousness isn't necessary to describe the physical world. Consciousness is necessary to fully describe a human being. Thus, consciousness must not be physical. Even if consciousness isn't necessary to fully describe a human being, it exists. Thus, consciousness must not be physical.)

I shall now go through the entire chain of logical possibilities. Any time I reach a possibility that proves physics isn't causally closed, then I will say so. At the end of the essay, I will have shown that every twig of the possibility tree terminates in causally open physics.

That is, while we can't yet say exactly why consciousness isn't physical, we can rule out the fact that it isn't. This is why I state this idea as a negation; I cannot describe consciousness. I can only determine a few things it is not.

I need a catchphrase so I can put it here. Something less lame than 'phear the logicz!' Let's begin.

First, let's assume you answered 'No, I'm not conscious. I do not experience qualia. Consciousness does not exist and the question is meaningless.' This is logically incoherent. I'm conscious. I can't be confused about my experience; it can't be some illusion; my experience of experience is consciousness itself. I think, therefore, I am.

So, are you conscious? You're welcome to answer no, but I'm taking all your stuff. If you're not conscious then you cannot suffer.

Next, did you evolve? If you didn't evolve, then this essay is moot anyway - we know that there's something beyond physics. If humans did not evolve, then biology is not causally closed under physics.

If you evolved and you are conscious, then consciousness is either free or confers some specific fitness advantage.

If consciousness is free, then that means it has no physical overhead. It appears spontaneously. It either has no physical effects in which case it is purely nonphysical, or it has physical effects and your brain is not causally closed under physics. (There would also be no constraints on the occurrence of consciousness.)

However, consciousness is not free. First, it requires specialized brain structures, and second consciousness has the property of 'choice' or 'decision' which allows you to, among other things, kill yourself.

Thus consciousness must confer a selective advantage that outweighs the costs. However, it is physically impossible.

First let me note that I'm distinguishing thoughts from actions; information processing from rearrangements of energy.

The relevant action, reproduction, is entirely physical, as are all the relevant prerequisites. However, all of these actions could be carried out without experiencing them. We could program, hard-script, a robot to do so without much trouble. (If computers are conscious then consciousness is free.)

Let me digress, considering some physical action that cannot be performed by a robot. What is preventing it? Are their hands not simply patterns of atoms, just as ours are? If we were to find a test of consciousness it would mean we have literally discovered magic. If there is an action robots cannot perform, then physics is not causally closed under physics.

Similarly, things like 'creativity' which are strings of fully physical actions, cannot be unique to consciousness. It would mean that there's literally an incantation that you can do to prove you're conscious. Which step exactly would a robot be unable to mimic and why? How would it know to fail if you hard-code it to mimic the action?

Further, the relevant thoughts are similarly constrained. There is no calculation that can be performed by a brain that cannot be performed by a computer. To posit such a calculation would require it to be not a calculation - it would require it to be non-mathematical and thus consciousness would perform nonphysical calculations.

As such, a robot can perform all the relevant physical actions of reproduction. Because computers have a much easier time staying consistent, it would likely perform the actions and calculations at a much higher efficiency than a human would.

Thus, a computer is fully capable of emulating human action, in the physical world, without consciousness. There is no physical action or thought that consciousness is a prerequisite for.

What if all this is true, but consciousness is physical but just somehow does it better than a computer? That, sure, you don't have to experience it, but experience makes it more effective? That would literally mean consciousness is a magic ingredient. Two otherwise identical computers compete, one experiencing its calculations, one not, and the conscious one wins. Consciousness would affect the physical world without actually being a part of it, contradicting the premise.

Consciousness physically does nothing. Consciousness confers a selection advantage. Your brain is causally open under physics.

This isn't easy, let's recap.
  1. You evolved.
  2. You are conscious.
  3. Consciousness costs resources.
  4. There is no action you must experience to perform.
  5. There is no action you could perform better by experiencing it.
  6. There is no computation you must experience to perform.
  7. Ergo, consciousness cannot physically do anything.
  8. Consciousness confers a selective advantage.
  9. Ergo, consciousness isn't physical.
  10. Your brain is causally open under physics.
  11. Ergo, the universe is causally open under physics.
And if any of the premises are untrue, then the conclusion remains true. Further and less obviously, there are no other possibilities to consider. It is logically impossible for me to be wrong.

Physics is best described as a consistent set of rules for the interaction of energy. Without such a set, interaction is in fact impossible, because contradictions do not exist.

Therefore, I propose that there is an orthogonal physics that uniquely defines the physical output of consciousness. I could call it the physics2 of vacuum2. (I won't.) This system must have a similar set of rules, though not necessarily mathematical ones. Also, if it is to not violate causality, (which you'll note I used to discover it) the system must be incapable of violating any physical law.

The system presumably has a similar causal hole. The system mutually closes causality with the physical vacuum.

This is the set of rules that allows us to be conscious - to experience the world. It is the plane of the mind. Unfortunately, that's nearly everything I know about it.

Most likely it allows free will, but it is possible that it simply confirms determinism. [rant]So please, quit debating it. You have no idea. The concepts of free will and determinism aren't even consistent concepts. Believe whatever the hell you want and tell anyone who argues with you to stuff it.[/rant]

So, in physics, what's the hole? There's only one hole in physics that I know of. It's been rigorously proven, too. Quantum randomness is true stochastic behavior. There's no physical way to predict what any particular particle will do, ever.

Technically speaking, a second set of rules could easily interface at this level, without violating any law or indeed being detectable at all, except through logic.

But for now I'm done with the firm certainty of deductive logic. This tree has grown up and borne delicious philosophical fruit. In my next essay I'll discuss some of my minor hobbies, the consequences of this fruit, and develop some of the ways consciousness may be implemented in the brain.