Comment Permalink

Consciousness is a roughly defined and (leaky) abstraction.

So this leads to the conclusion that, if a Turing machine computes consciousness and summarizes its output in a static representation on a tape, the tape is conscious.

Without context the content of the tape has no meaning. So the consciousness that has been output on the tape, is only a consciousness in the context that can use it to generate the consciousness abstraction.

It is the set of "stuff" that produces the consciousness abstraction that can be called conscious. In a Turing machine, this "stuff" would be the tape plus the machine that gives the tape the necessary context.

PhilGoetz14y00

As Nisan asked above: Is this Turing machine conscious if you don't run it?

See in context

16 The conscious tape

by PhilGoetz

16th Sep 2010

4 min read

119

16

This post comprises one question and no answers. You have been warned.

I was reading "How minds can be computational systems", by William Rapaport, and something caught my attention. He wrote,

Computationalism is - or ought to be - the thesis that cognition is computable ... Note, first, that I have said that computationalism is the thesis that cognition is computable, not that it is computation (as Pylyshyn 1985 p. xiii characterizes it). ... To say that cognition is computable is to say that there is an algorithm - more likely, a collection of interrelated algorithms - that computes it. So, what does it mean to say that something 'computes cognition'? ... cognition is computable if and only if there is an algorithm ... that computes this function (or functions).

Rapaport was talking about cognition, not consciousness. The contention between these hypothesis is, however, only interesting if you are talking about consciousness; if you're talking about "cognition", it's just a choice between two different ways to define cognition.

When it comes to consciousness, I consider myself a computationalist. But I hadn't realized before that my explanation of consciousness as computational "works" by jumping back and forth between those two incompatible positions. Each one provides part of what I need; but each, on its own, seems impossible to me; and they are probably mutually exclusive.

Option 1: Consciousness is computed

If consciousness is computed, then there are no necessary dynamics. All that matters is getting the right output. It doesn't matter what algorithm you use to get that output, or what physical machinery you use to compute it. In the real world, it matters how fast you compute it; but surely you can provide a simulated world at the right speed for your slow or fast algorithm. In humans today, the output is not produced all at once - but from a computationalist perspective, that isn't important. I know "emergence" is wonderful, but it's still Turing-computable. Whatever a "correct" sequence of inputs and outputs is, even if they overlap in time, you can summarize the inputs over time in a single static representation, and the outputs in a static representation.

So what is conscious, in this view? Well, the algorithm doesn't matter - remember, we're not asking for O(consciousness); we're saying that consciousness is computed, and therefore is the output of a computation. The machine doing the computing is one step further removed than the algorithm, so it's certainly not eligible as the seat of consciousness; it can be replaced by an infinite number of computationally-equivalent different substrates.

Whatever it is that's conscious, you can compute it and represent it in a static form. The simplest interpretation is that the output itself is conscious. So this leads to the conclusion that, if a Turing machine computes consciousness and summarizes its output in a static representation on a tape, the tape is conscious. Or the information on the tape, or - whatever it is that's conscious, it is a static thing, not a living, dynamic thing. If computation is an output, process doesn't matter. Time doesn't enter into it.

The only way out of this is to claim that an output that, when coming out of a dynamic real-time system, is conscious, becomes unconscious when it's converted into a static representation, even if the two representations contain exactly the same information. (X and Y have the same information if an observer can translate X into Y, and Y into X. The requirement for an observer may be problematic here.) This strikes me as not being computationalist at all. Computationalism means considering two computational outputs equivalent if they contain the same information, whether they're computed with neurons and represented as membrane potentials, or computed with Tinkertoys and represented by rotations of a set of wheels. Is the syntactic transformation from a dynamic to a static representation a greater qualitative change than the transformation from tinkertoys to neurons? I don't think so.

Option 2: Consciousness is computation

If consciousness is computation, then we have the satisfying feeling that how we do those computations matters. But then we're not computationalists anymore!

A computational analysis will never say that one algorithm for producing a series of outputs produces an extra computational effect (consciousness) that another method does not. If it's not output, or internal representational state, it doesn't count. There are no other "by-products of computation". If you use a context-sensitive grammar to match a regular expression, it doesn't make the answer more special than if you used a regular grammar.

Don't protest that a human talks and walks and thereby produces side-effects during the computation. That is not a computational analysis. A computational analysis will give the same result if you translate whatever the algorithm and machine running it is, onto tape in a Turing machine. Anything that gives a different result is not a computational analysis. If these side-effects don't show up on the tape, it's because you forgot to represent them.

An analysis of the actual computation process, as opposed to its output, could be a thermodynamic analysis, which would care about things like how many bits the algorithm erased internally. I find it hard to believe that consciousness is a particular pattern of entropy production or waste heat. Or it could be a complexity or runtime analysis, that cared about how long it took. A complexity analysis has a categorical output; there's no such thing as a function being "a little bit recursively enumerable", as I believe there is with consciousness. So I'd be surprised if "conscious" is a property of an algorithm in the same way that "recursively enumerable" is. A runtime analysis can give more quantitative answers, but I'm pretty sure you can't become conscious by increasing your runtime. (Otherwise, Windows Vista would be conscious.)

Option 3: Consciousness is the result of quantum effects in microtubules

Just kidding. Option 3 is left as an exercise for the reader, because I'm stuck. I think a promising angle to pursue would be the necessity of an external observer to interpret the "conscious tape". Perhaps a conscious computational device is one that observes itself and provides its own semantics. I don't understand how any process can do that; but a static representation clearly can't.

ADDED

Many people are replying by saying, "Obviously, option 2 is correct," then listing arguments for, without addressing the problems with option 2. That's cheating.

Personal Blog

16

New Comment

119 comments, sorted by

top scoring

Click to highlight new comments since: Today at 1:31 AM

Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings

[-]Jonii14y60

Or the information on the tape, or - whatever it is that's conscious, it is a static thing, not a living, dynamic thing. If computation is an output, process doesn't matter. Time doesn't enter into it.

Time is right where it should be in at: In the tape. We, the Matrix-Lords, are above the time of the tape, but it doesn't mean the consciousness within the tape isn't living and dynamic and all that.

The question presented is very intriguing, thank you for it.

[-]Nisan14y50

This question arises when I consider the moral status of intelligent agents. If I encounter a morally-significant dormant Turing machine with no input devices, do I need to turn it on?

If yes, notice that state N of the machine can be encoded as the initial state of the machine plus the number N. Would it suffice to just start incrementing a counter and say that the machine is running?

If I do not need to turn anything on, I might as well destroy the machine, because the Turing machine will still exist in a Platonic sense, and the Platonic machine won't noti... (read more)

3jacob_cannell14y

For your moral questions, I think it would help if you replace "morally significant dormant turing machine with no input devices" with "comatose human". Notice that the state N of a comatose human patient can be encoded as the initial state plus the number N. Would it suffice to just start incrementing a stopwatch and say that the patient is well? If I do not need to turn anything on, I might as well destroy the patient, because the patient will still exist in a Platonic sense, and the Platonic patient won't notice if I destroy a manifestation of it. The only platonic sense in which things still exist after being destroyed is in the sense of us remembering and thinking about them - a very weak form of simulation. If we could think in much more precision and with vastly more power, then we could make thought-things 'real'. But until we have simulations of such power, all we have is the real world. And nonetheless, everything that exists, even in simulation, must be encoded somewhere with matter/energy in the universe.

1Nisan14y

Ah, but presumably if we were to wake up the comatose person, they would start interacting with the world; and their output would depend on the particulars of the state of the world. In that case I clearly want to wake them up. I was thinking of a morally significant dormant Turing machine that was not designed to have input devices. For example, a comatose person with no sensory organs. If they woke up, they would awaken to a life of dreams and dark solitude, proceeding deterministically from their initial state. Let's assume there is absolutely no way to restore this person's senses. It's not clear to me that it's morally desirable to wake them up.

2David_Allen14y

Good summary. Yes my statements are in part a recasting of the functionalism philosophy mentioned by Jacob Cannell, in terms of the context principle, which I describe here.

1Armok_GoB14y

If this is an ACTUAL situation as described, rather than the contrived one you intended, you should copy the contents to somewhere you have good control over, then run it and meddle with it to give it I/O devices, or run it for as far as the agent(s) in it would have wanted it to run and then add I/O devices, or extract the agents as citizens in your FAI optimized place to have fun, or something along those lines like.

1MatthewW14y

It seems to me that the arguments so lucidly presented elsewhere on Less Wrong would say that the machine is conscious whether or not it is run, and indeed whether or not it is built in the first place: if the Turing machine outputs a philosophical paper on the question of consciousness of the same kind that human philosophers write, we're supposed to take it as conscious.

1lukstafi14y

It is useful to distinguish the properties "a subsystem C of X is conscious in X" and "C exists in a conscious way" (which means that additionally X=reality). I think Nisan expresses that idea in the parent comment.

0[anonymous]14y

The machine considered has the property of being conscious in its context X (i.e. X = the system containing the machine, the producers of its input and consumers of its output). The machine exists in a conscious way if additionally X = reality.

[-]DanielVarga14y40

ADDED: Many people are replying by saying, "Obviously, option 2 is correct," then listing arguments for, without addressing the problems with option 2. That's cheating.

Phil, I have to say that I don't think the problems with option 2 are actually presented in you post. But that does not mean that we are allowed to dodge the question implicit in your post: how to formally distinguish between two computational processes, one conscious, the other not. Let me start my attempt with a quote:

"Consciousness is overrated. What we call consciou

... (read more)

[-]DanArmak14y40

What is consciousness, and what kinds of things are conscious?

I've seen this debated many times, and I suspect that people are merely arguing over the meaning of a word that does not happen to carve reality at the joints.

What is it in the physical universe that you call consciousness and which might be present or absent in a computational device? What prediction is being made by any of the theories 1 through 3 in this post, that I can go out and test?

[-]orthonormal14y40

I endorse the first alternative; the intuition at first felt wrong (in a Chinese Room sort of way), but that feeling disappeared when I realized the following:

I was envisioning a tape (call it Tape A) which only recorded some very small end result of Turing Machine A, like the numerical output of a calculation or the move Deep Blue makes. And that seems too "small" somehow to encapsulate consciousness— I felt that I needed the moving Turing machine to make it "live" in all its detail.

But of course, it's trivial to write a different Turi... (read more)

0David_Allen14y

In what context can Tape B be labeled conscious? A history of consciousness does not seem to me to be the same as consciousness. A full debug trace of a program is simply not the same thing as the original program. If however you create a Machine C that replays Tape B, I would grant that Machine C reproduces the consciousness of Machine A.

1orthonormal14y

This gets into hairy territory with no clear "conscious"/"not conscious" boundary between a spectrum of different variations, but I'd say that the interpretive framework needed to trace a thought from the log on Tape B is essentially the same as the interpretive framework needed to trace it from the action of Machine A on the start tape. They're isomorphic mathematical objects.

0David_Allen14y

I agree with everything you say here. I claim that the "interpretive framework" you refer to is essential in the labeling of Tape B as conscious. Without specifying the context, the consciousness of Tape B is unknown.

5orthonormal14y

You might be interested in the thought experiment of a so-called "joke interpretation", which maps the random molecule oscillations in (say) a rock onto a conscious mind, and asks what the difference is between this and a more "reasonable" map from a brain to a mind. There's a good discussion of this in Good and Real.

0David_Allen14y

I skimmed the material and see what you mean. I would restate the thought experiment as such. A state sequence measured from a rock is used to generate a look-up table that maps from the rock state sequence to a pre-measured consciousness state sequence. This is essentially an encryption of the consciousness state sequence using the rock state sequence as a one-time pad. The consciousness state sequence can be generated by replaying the rock state sequence through the look-up table. With the final question being: is the rock conscious? In the model I've outlined in my comments, consciousness exists at the level the consciousness abstraction is present. In this case that abstraction is not present at the level of the rock, but only at the level of the system that uses the look-up table, and only for the duration of the sequence. The states measured from the rock are used to generate the consciousness, but they are not the consciousness.

1bogus14y

What is the "system that uses the look-up table"? Do you require a particular kind of physical system in order for consciousness to exist? If not, what if the "system" which replays the sequence is a human with a pen and paper? Does the system truly exhibit the original consciousness sequence, in addition to the human's existing consciousness?

1David_Allen14y

Ah, Chinese room questions. The system that replays the sequence can be anything, including a human with pen and paper. Yes, assuming that the measured consciousness sequence captured the essential elements of the original consciousness. To "see" the original consciousness in this system you must adopt the correct context; the context that resolves the consciousness abstraction within the system. From that context you will not see the human. If you see a human following instructions and making notes, you will not see the consciousness he is generating. Consider a chess program playing a game against itself. If we glance at the monitor we would see the game as it progresses. If instead we only could examine the quarks that make up the computer, we would be completely blind to the chess program abstraction.

2PhilGoetz14y

Is a mono-consciousness then impossible?

0David_Allen14y

Thanks for the clarification. In my comments I have been working on the idea that consciousness is an abstraction. The context in which the consciousness abstraction exists, is where consciousness can be found. So a mono-consciousness would still have a context that supports a consciousness abstraction. I don't see any problem with that. However the consciousness might be like a feral child, no table manners and very strange to us. How about this. If a consciousness tells a joke in a forest where no other consciousness can hear it, is the joke still funny?

0David_Allen14y

I'll need more details. What is a mono-consciousness?

0PhilGoetz14y

I was thinking of a magnetic monopole. A single consciousness, that does not interact with any others

[-]Mass_Driver14y40

Your question is well-posed, but I doubt that it really attacks the problem of consciousness.

I don't understand what it could possibly mean for an output or an algorithm to be consciousness. Consciousness, whatever it might be caused by or composed of, means subjective awareness of qualia.

In this sense, "3" is not even slightly conscious. Neither is "output=0; if input=3, let output=1." Neither is "output=0; if input=3, let output=1, get input by scanning source code of self and reducing it to a number." The last example will ... (read more)

3jacob_cannell14y

Everything that exists can be described precisely by some physics or algorithm, down to the point where it's actually meaningless to differentiate between the algorithm and the process itself. If consciousness exists, there exists some algorithm that is exactly equivalent to it. Saying the only response to any questions about consciousness is "I don't know" is equivalent to shielding consciousness with the holy eternal veil of mystery. Within the fields of computational neuroscience and AI, we actually do know a great deal about the algorithms underlying the supposed mystery of 'consciousness'. We dont' know enough yet to engineer a conscious machine, but that isn't so far away. We had a general idea of how the heart worked long before we made an artificial heart, and we are in a similar situation with the brain. But the brain happens to be more complex than the heart.

5Mass_Driver14y

How do you know?

5orthonormal14y

Because that thesis has made better predictions than every rival theory which seemed at the time more reasonable (superstition, vitalism, Cartesian dualism, etc). The Pythagoreans, amidst all their lunacy, stated perhaps the world's best-confirmed audacious hypothesis, that the world is a mathematical object.

5kodos9614y

When it comes to the question of consciousness, I humbly submit that "i-don't-know-ism" has made better predictions (i.e. none) than any rival theory.

3orthonormal14y

The thesis doesn't predict that every reduction is going to be easy. And "I don't know" really masks a good bit of knowledge, unless you're equally surprised by all new data. Physicalism directly predicts a good many of the things we consider too obvious to categorize as "mysterious" (e.g. that brain damage can cause personality change).

0Mass_Driver14y

Sure! And I would take decent odds on physicalism...at 20:1 in my favor I wouldn't have to think too hard; I'd be pretty sure to take the bet, because there is some pretty convincing evidence, like how brain damage works and how simple formulas about e.g. mechanics or radiation explain phenomena in what we might naively assume to be different realms, e.g. solar sails and roof albedo and warm light bulbs. If you can formulate a hypothesis with one set of data and test it using several other sets and get confirmation, it makes sense to guess that it works on all sets. And if you put a gun to my head and said "guess a theory of everything," I'd guess physicalism...as I said earlier, "of course physicalism or whatever you want to call it is the most plausible known and articulated theory of everything." My only point is that we don't yet have enough evidence to be sure of physicalism in its broadest senses so as to justify shutting down alternative avenues of exploration for standing questions such as the origin of the universe, the nature of consciousness, and the computability of matter.

1Mass_Driver14y

The results of science are indeed quite impressive. Suppose you wanted to compare the Pythagorean hypothesis "the world is a mathematical object" with the slightly broader hypothesis "the world consists largely of objects following mathematical laws." Are there scientific results that would be predicted by one hypothesis but not the other?

1orthonormal14y

No– this is a variant of the "green/grue" problem. However, the Pythagorean hypothesis puts higher probability on the things we've actually observed, because it doesn't waste any on claiming that this thing or that is non-mathematical. By Bayes' Law, this means it's continually gaining support against the rival candidate.

1Mass_Driver14y

Would you be so kind as to define "mathematical object"? Possibly I agree with you on everything but semantics, a field in which I am almost always happy to compromise.

1orthonormal14y

Er, a set with a simple definition, like the Mandelbrot set or the set of solutions to the Schrödinger equation on a given manifold? Honestly, I'd be surprised if this is the point you're stuck on. What I suspect might help is the distinction here between epistemology and ontology: it's a meaningful hypothesis that we live in such a mathematical object, even if there doesn't exist a mind sufficient to exhaustively verify this, and yet our smaller minds can acquire enough evidence about the world's structure to raise that hypothesis to near certainty (modulo some chance of being in a simulation that's more complicated than the laws we seek, but whose creators want us not to notice the seams).

0Mass_Driver14y

I think you're right that what we disagree about is The dichotomy you've provided seems to me to be an excellent definition of the difference between mathematical epistemological proof and empirical epistemological proof...it happens all the time that we may not be able to rigorously show N, but we nevertheless have extremely good reason to believe N with near-certainty, and even stronger reason to act as if we believed N. If I hear you correctly, you think that we could plug in "the Universe is merely a mathematical object" for N. I disagree. For me, the difference between epistemology and ontology is that there is a difference between what we can know and what exists. There might be things that exist about which we know nothing. There could even be things that exist about which we cannot know anything. One could reasonably call for scientists to ignore all such hypothetical objects, but, philosophically speaking, it doesn't stop the objects from existing. It boggles my mind to hear the claim that a mathematical object, as you have just defined it in your last comment, "exists in this second, ontological sense. The mandelbrot set expresses a relationship among points. If several small spheres exist and it turns out that the points approximate the relationship defined by the Mandlebrot set, then we might say that a Mandlebrot-ish shape of spheres exists. But the set itself doesn't have any independent existence. This result doesn't seem to me to depend on whether we use spheres or rays or standing waves -- you still have to be vibrating something* if you want to talk about things that actually exist. I'm not the sort of nut that believes in good old-fashioned aether, but mathematical relationships alone won't get you a flesh-and-blood universe where things actually exist...they'll just get you a blueprint for one. Even if, epistemologically, we can know everything about the blueprint and model all of its parameters, it still won't exist unless it's made of somet

8orthonormal14y

It continually amazes me that people think "physical existence" is somehow less mysterious and more fundamental than the existence of a mathematical object!

0Mass_Driver14y

Er, no, it's not less mysterious -- we understand mathematical objects better than we understand physical existence; mathematical objects can be treated with, well, math, and physical existence gets dealt with by jokes like philosophy. I'm not sure what you mean by more fundamental, but physical existence does seem to be roughly as important as mathematical objects...at any rate, it matters a lot to me whether things exist in fact or merely in theory.

0orthonormal14y

We've been given special evidence in our own case, but if we step away from that for a moment, what I mean should be clear. Let's take a hypothetical Universe X, which is very different from ours. Saying "Universe X is a simple mathematical object" is pretty well comprehensible. Saying "Universe X exists in some special way, distinct from just being a mathematical object, and in fact it might not be describable as a simple mathematical object" is just plain mysterious. It's up for debate whether it's even a meaningful statement. Apply that to discussion of our own universe.

1Will_Newsome14y

But, but, you don't understand. Math isn't reeeeeeeaaaaaaaaaaaaaal! I have an idea. Maybe it'd be more convincing if you said "Universe X is a simple computation." People feel like computations are more real, and who knows, maybe they're right. Maybe reality is computation, just a subset of mathematics. It seems a lot easier for people to envision that, at any rate. Or take Eliezer who (I think?) seems to think (or at least seemed to think) that reality juice is magically related to acyclic causal graphs. You'll still probably get the same objections, though: "Computations aren't reeeeeeeaaaaaaaal, they have to be computed on something! Where's the something coming from?" But that seems a little bit more silly, because the Something that is computing can be infinitely far back in the chain of computation. All of a sudden it feels more arbitrary to be postulating a Something that is Real. And real metaphysicists know that things shouldn't feel arbitrary.

0gwern14y

Now you're just ripping off the last chapter of Drescher's Good and Real. You know he comments here sometimes - he'd be so hurt at such plagiarism. :)

0Will_Newsome14y

I am? I only read the decision theory chapters of Good and Real, the day before he showed up at SIAI house for the decision theory workshop. I'll definitely read the last chapter when I get back to California.

1Will_Newsome14y

I think your intuition is relying a little too much on the absurdity heuristic (e.g., "It boggles my mind...") and flat out assertion (e.g., "But the set itself doesn't have any independent existence."). Metaphysical intuition is really misleading. I think most people underestimate that, especially because the absurdity heuristic is strong and therefore it's easy to reach a reductio ad absurdum that is nonetheless true. I'll give an example. Once upon a time I didn't think copies 'counted' in a multiverse, either morally or for purposes of anthropic reasoning. 200 Jacks had the same weight as 1 Mary. The opposite was absurd, you see: You're claiming that 3 copies of the exact same computation are worth more than 2 computations of 2 different people, leading separate and diverse lives? Absurd! My moral and metaphysical intuition balks at such an idea! I came up with, like, 3 reductio ad absurdums to prove my point. Eliezer, Wei Dai, Steven Kaas, Nick Bostrom, what did they know? And there was some pride, too, because they way I was thinking about it meant I could easily deal with indexical uncertainty, and the others seemed clueless. ... Well, turns out those reductios weren't absurd: I just hadn't learned to think like reality. I had to update, because that's where the decision theory led, and it's hard to argue with mathematics. And it came to my attention that thinking doubled computations had the same measure had a lot of problems as well. Since then, I've been a lot more careful about asserting my intuition when it disagrees with people who seem to have thought about it a lot more than I have. In the case of the Mathematical Universe Hypothesis or permutations thereof (Eliezer seems to think the mysterious 'reality fluid' or 'measure' has a lot to do with directed acyclic graphs, for instance), there's a lot of mental firepower aimed against you. Why do you believe what you believe? If it turns out the reason is metaphysical intuition, be on guard. Acknowledge

1Mass_Driver14y

Look, of course physicalism or whatever you want to call it is the most plausible known and articulated theory of everything. But why would you assign physicalism nontrivial probability as against (a) theories that are as yet unknown or unarticulated, or (b) the possibility that the Universe does not behave neatly in accordance with a single coherent, comprehensible theory? Isn't the concept-space of "single coherent Theory of Everything" vastly smaller than the total concept-space of concepts that could describe our reality?

6orthonormal14y

The thesis at hand predicts that we should find complex things to be intricate arrangements of simple things, acting according to mathematically simple rules. We have discovered this to be true to a staggering degree, and to the immense surprise of the intellectual tradition of Planet Earth. (I mean, when even Nietzsche acknowledges this— I'll reply later with the quote— that's saying something!) Your (b) makes no such specific predictions, and so the likelihood ratio should now be immensely in physicalism's favor. Only a ridiculous prior could make it respectable at the moment. As for (a), I'm talking about the general principle that the world is a mathematical object, not any particular claim of which object it is. (If I knew that, I'd go down and taunt the string theorists all evening.)

2orthonormal14y

* Nietzsche, The Gay Science I.46 (NB: in this passage, "we" signifies modern atheists, not people in general.)

0[anonymous]14y

As opposed to what? Two or more incoherent theories? Isn't that just a strange way to talk about an impossible reality?

-1jacob_cannell14y

In short, because physics is so successful. In long, because no matter how far off physics is from the ultimate algorithm, we can continue to narrow in on it indefinetly. Mathematically at least, even an infinite algorithm is possible. As a curious side note, I remember physcist Frank Tipler has a GUT of physics that is infinite. He claims this TOE has been known for a while, but avoided for obvious reasons. He then puts on a magic space cap and claims that this TOE proves Christianity is correct, but the TOE is interesting nonetheless (at least the idea of it - I am not a physicist). I don't know for certain that physics is computable, but from what I have read on that matter, all current indications are positive.

1Mass_Driver14y

Successful at what, exactly? At modeling the behavior of the stuff that humans can easily observe using basic industrial technology over the span of 100 to 400 years? Why would you want to extrapolate from that to "everything that exists?" Right, but what makes you think there is an "ultimate algorithm" to be found?

2jacob_cannell14y

A single universal physics is adequate to explain all that we can observe, and a necessary derivation of that universal physics is a vast quantity of space and time which we can not directly observe but which we predict is also driven by the same universal physics. This is the "everything that exists" - whose existence is in some fact dependent on the universal physics itself. Whether there is or is not an ultimate algorithm is not even the right question. It is true by default. We can continue to refine physics indefinetly. In other words, of course there is an ultimate algorithm, because we can invent it. In fact, given any sequence of finite observations O, there is an infinite set of algorithms A that perfectly predict/compute the sequence O. Physics is concerned largely with finding the minimally complex algorithm that fully predicts O. So yes, mathematically it is trivially true that there is an infinite set of ultimate algorithms.

1Mass_Driver14y

Thank you for one (of several) intelligent responses. This isn't quite right. The only thing that makes the derivation "necessary" is your adjective "universal." We could just as easily say that there is a supergalactic physics that explains all we can observe, and that same physics could plausibly explain what is happening in the space and time that we cannot or have not observed. Note that the unobservable realms are not merely those outside our past light cone, but also those within the limits of the Heisenberg uncertainty principle, beneath the smallest structures that we can repeatedly observe, and, for all practical purposes, the space beyond the nearest nebula and/or the objects too dull for our Earth-bound telescopes to detect. It would be remarkably bad science to voluntarily choose to sample only one kilobyte from one address out of thousands of terabytes of data and assume that the kilobyte is representative. The fact that all known scientific resources are clustered in the same tiny portion of spacetime forces us to use such a sample, but it cannot and should not force us to assume that the sample is representative. I don't understand what you mean. Intelligent minds with an ability to manipulate matter or energy can 'create' patterns in that matter/energy by rearranging it according to the laws of physics. However, I cannot think of any sense in which physics itself could be said to create its own patterns. Physics is the pattern in which all known matter is currently arranged, but physics does not create the matter -- it merely arranges it. Physics does not explain why there is something instead of nothing; it would be perfectly consistent with the laws of physics for there to be no electrons orbiting no protons over a volume of no space-time. How then can "everything that exists" be dependent on physics? Right, but who says our observations are finite? What if important phenomenon, like, e.g., consciousness (cough) turn out to depend on infinitely

4jacob_cannell14y

Physics is generally held to be universal, instead of just 'supergalactic'. For one, there is the multiverse. But in general, the idea is, as I discuss later, to find the most parsimonious explanation for everything. This is the optimal strategy, and universality is a necessary consequence of this strategy. Any other physics or system which does not explain all observations is of course incomplete and inferior. Not at all. You seem to be applying the analogy that at the cosmic scale the universe is some sort of probabilistic urn that generates galactic-sized space-time slices at random whim. It is not. There are an infinite set of potential physics that have widely different properties in regions we can not observe. There are strong reasons why these are all necessarily inferior, by the principle of Ockham's razor and the low-complexity bias in Solonomoff induction. Elementary physics. There are a finite number of humans, the earth has finite mass, finite information storage potential, and we have finite knowledge. If you want to believe something like this is true before you begin, that consciousness is somehow different and special, then you are abandoning rationality from the start. There are no privileged hypothesizes and no predefined targets in the quest for knowledge. Sure, infinities are possible, although they generally are viewed to signal a problem in physics when they come up in one's math. But that's all besides the point: our observations are obviously finite. And furthermore, infinities are not at all an obstacle towards a universal physics. There is no such complexity limit whatsoever to computability - it is not as if a phenomena has to be sufficiently 'simple' for it to be computable in theory (although practical computability is a more complex issue). True randomness comes up immediately in quantum mechanics. This isn't an obstacle to computability, whether theoretical or practical. People unfamiliar with computing often have the notion t

-1Mass_Driver14y

OK, thank you for talking with me. I've lost interest in the conversation, partly because of your minor ad hominem attack ("sounds to me like you have a mystery that you would like to protect"), but mostly because I see your arguments as dependent on assumptions that I do not share: you see it as "obvious" that physics is universal and that theories favored by Solonomoff simplicity are automatically and lexically superior to all other theories, and I do not. If you care to defend or explain these assumptions, I might regain interest, or I might not. Proceed at your own risk of wasting your time. In any case, thank you for a stimulating debate.

1jacob_cannell14y

Sorry for the minor ad hominem, I jumped to the conclusion based on prior experience. We can generate explanations for anything. Science has found that the universe appears to operate on a universal set of underlying principles - everything reduces to physics. We could have lived in a universe where this wasn't so. But we don't. The computer I am working on right now is solid proof of physic's success. When you have two theories (algorithms) that both accurately prediction an observation sequence, you need some other criteria to guide you - and here ockham's razor comes in to play. There are always an infinite number of more complex theories that explain a series of observations, but only one that is minimally simple. But again, I think the universality of physics stems just from the simple fact that there are an infinite number of algorithms (theories) that can explain any possible sequence of observations - so universality is always possible.

0Tiiba14y

"Why would you want to extrapolate from that to "everything that exists?"" That's all we've got?

3Vladimir_M14y

jacob_cannell: What exactly is an "algorithm," according to your usage of the term?

0torekp14y

I too would like the definition of "algorithm" here, with attention to the difference (Tegmark be damned) between mathematical objects and physical objects/processes.

[-]jimrandomh14y40

In humans today, the output is not produced all at once -but from a computationalist perspective, that isn't important.

I don't think this is valid; the complexity you're discarding here may actually be essential. In particular, interacting with some sort of universe that provides inputs and responds to outputs may be a necessary condition for consciousness. Perhaps consciousness ought to be a two-place predicate. If you have a universe U1 containing a mind M, and you simulate the physics of U1 and save the final state on a tape that exists in universe U2, then conscious(M,U1) but ~conscious(M,U2). On the other hand, if U1 interacts with U2 while it's running, then conscious(M,U2).

1PhilGoetz14y

Is the universe not something that can be represented as information? Do you mean U1 includes the specific materials used for its reality? That would be taking Searle's "conscious brains must be made of conscious-brain stuff" argument, and changing it to "conscious brains must be surrounded by consciousness-inducing universe stuff".

0jacob_cannell14y

The universe can be represented as computation, but it appears that requires a time element. You can not define a turing machine based on just a tape - it intrinsically requires a time dynamic in the form of the moving head. So in digital physics and computationalism, time is fundemental - it really exists and can not be abstracted away. At the most core the universe is something that changes. Described as a turing machine it consists of information (the tape) and time - the mover.

1PhilGoetz14y

That sound right - but that leads to Option 2.

0jacob_cannell14y

Yes your option 2 sounds almost right, but see my amendments/disagreements.

0DanArmak14y

A TM has an infinite working tape, but the problem of human-like consciousness has only finite input and output sizes. Therefore, you could define an (infinite) function by simply listing all possible pairs of (finite) inputs and outputs. This is a completely static, time-less representation that is still powerful enough to compute anything with bounded inputs and outputs.

2jacob_cannell14y

Of course you can collapse any function into a static precomputation, but that is still itself a function, and it still requires at least one step, and it still requires the turing machine head, so you have not removed time. I'm aware of no time-less representation of a turing machine, it seems impossible in principle. Furthermore, for the system to exist in the real world, it will have to produce outputs for particular inputs at particular times - the time requirement is also imposed by the fact of time in our universe.

0DanArmak14y

What you say is true. I find myself unsure how it applies to the original subject. Possibly my comment wasn't on topic... So feel free to ignore it.

[-]Will_Sawin14y30

Suppose we view consciousness as both a specific type of computation and a specific range of computable functions. For any N, there will always be a lookup table that appears conscious for a length of time N, in particular, "the lifetime of the conscious creature being simulated". A lookup table, as Eliezer once argued, is more like a cellphone than a person - it must have been copied off of some sort of real, conscious entity.

Is the feature of a lookup table that makes it unconscious its improbability without certain types of computation, that ... (read more)

6Vladimir_M14y

However, in physical reality, there exist no Turing machines. There are only finite state machines whose behavior emulates a quasi-Turing machine with a finite tape (or an analogous crippled finite version of some other Turing-complete theoretical construct). Now, every finite-state machine can be implemented using a lookup table and a transition function that simply performs a lookup based on the current state and input. Any computers we have now or in the future can be only clever optimizations of this model. For example, von Neumann machines (i.e. computers as we know them) avoid the impossibly large lookup table by implementing the transition function in the form of a processor that, at each step, examines one small subset of the state and produces a new state that differs only by another small subset based on simple rules. (I'm describing the effect of a single machine instruction, of course.) So, the question is: what exactly makes a lookup table deficient compared to a "real" computer, whatever that might be?

0PhilGoetz14y

Ouch. That hurts. This may be a better way to state the problem, because it doesn't intersect with the mysteries of time vs. static, and needing an observer.

1Vladimir_M14y

You might also be interested in this recent comment of mine, if you haven't read it already: http://lesswrong.com/lw/2m8/consciousness_of_simulations_uploads_a_reductio/2hky

[-]Paul Crowley14y20

A powerful computer in a sealed box is about to be fired away from the Earth at the speed of light; it will never produce output and we'll never see it again. From the point of view of perspective 1, the whole program is thus equivalent to a gigantic no-op. Nonetheless, I'd rather that the program running on it simulated conscious beings in Utopia than conscious beings in Hell. This I think forces me to perspective 2: that actually doing the calculations makes a moral difference.

EDIT: the "speed of light" thing was a mistake. Make that "close to the speed of light".

2Vladimir_M14y

I don't think your thought experiment is logically consistent. You're using a physical theory, namely special relativity, to discuss a case in which the theory explicitly refuses to say what happens, because it's considered unphysical within that theory. If the computer moves at exactly the speed of light, and assuming special relativity, the time in which the computer will reach a given step of its program becomes undefined, not "never." In any physically possible case, in which the computer's speed can be arbitrarily close to c, things develop completely normally in the computer's own reference frame. Moreover, if you observe the code of a program, it's just a string of bits (assuming a binary computer). And a string of bits can be interpreted as implementing any arbitrary program, given an arbitrary choice of the interpreter. Therefore, until an actual interpretation happens, what makes your hypothetical "hell" and "Utopia" essentially different?

3Paul Crowley14y

The "speed of light" qualification was a mistake. I was just trying to get the computer out of our light cone somewhere we can never observe it.

1Nisan14y

You could discriminate between heaven and hell by considering the minimum length of an interpreting program. Such a program would have to produce output that would be directly comprehendible by us, and it would have to be written in a language that we wouldn't regard as crazy. In order to see heaven in hell, your interpreter probably has to contain hell.

1Vladimir_M14y

Fair enough; that's true when it comes to an arbitrary program. However, consider a program that contains both heaven and hell in different branches, and will take one of these different branches depending on the interpreter. Or, alternatively, consider a program simulating a "good" world that will, given some small tweak in the orignal interpreter, simulate a much worse world because some simple but essential thing will be off. Such thought experiments, as far as I see, override this objection.

0wedrifid14y

For the right right value of 'arbitrary' the computer never performs a single operation. The entire box is obliterated by collision with a stray electron before the processor can tick. The collision releases arbitrarily large amounts of energy and from there things just start getting messy.

2Vladimir_M14y

In a thought experiment, you can assume anything, however unrealistic, as long as it's logically consistent with the theory on which you're basing it. Assuming away stray electrons is therefore OK in this particular thought experiment, since the assumption of a universe that would provide an endless completely obstacle-free path would still be consistent with special relativity. In fact, among the standard conventions for discussing thought experiments is not to bring up objections about such things, since it's presumed that the author is intentionally assuming them away to make a more essential point about something else. In contrast, introducing objects that move at exactly the speed c into a thought experiment based on special relativity results in a logical inconsistency. It's the same mistake as if you assumed that Peano axioms hold and then started talking about a natural number such that zero is its successor. Since the very definition of such an object involves a logical contradiction, nothing useful can ever come out of such a discussion.

0[anonymous]14y

Could someone please explain why this was downvoted? (I don't care about losing score, but I am concerned about the possibility that I wrote something stupid that I'm unaware of.)

-1[anonymous]14y

Of course it is OK. But since it was unspecified it was a whole lot more interesting to imagine the effects of a cataclysmic collision with arbitrarily large energy. (Because any discussion of a question of consciousness that goes for more than 3 paragraphs before dissolving the question or finding an interesting tangent is at least two and a half paragraphs too long!) Now I'm wondering whether such a collision would release enough light to obliterate Earth from an arbitrarily large (but within light cone) distance away. I'm thinking it would.

0PhilGoetz14y

The speed of light qualification is interesting, because it may relate to the static aspect of the "conscious tape". The computer is conscious in its reference frame; but since that clock is stopped, that consciousness will never begin.

1Paul Crowley14y

As I observe to Vladimir_M, the "speed of light" thing was a mistake. I just wanted to make sure no-one ever observed any output from the computer under any circumstances.

4wedrifid14y

You could always just hide it in the forest near the oft-considered fallen tree.

0PhilGoetz14y

No, I still think it's interesting. Define time as a function of entropy (eg., "one second" means the time over which entropy increases by a constant amount). Time is stopped in the piece of paper's reference frame, because the paper is static, and therefore has no entropy change, and therefore no passage of time.

[-]Kaj_Sotala14y20

I don't understand the difference between "computed" and "computation", here.

[-]Snowyowl14y100

"Computed" means that only the input and output are important: as long as you can get from "2+2" to "4", it doesn't matter how you do it. "Computation" means that it's the algorithm you use that is important.

If a computer can give the response a human would have to a given situation, despite that computer using an AI which operates on different principles from the human brain (simulating a universe containing a human brain is sufficient), is that computer thinking/conscious? If yes, then thought/consciousness can be computed. If no, then thought/consciousness is the computation.

This is related to the Turing Test, in which a computer is deemed conscious if it can produce responses indistinguishable from those of a human, regardless of the algorithm used.

0Kaj_Sotala14y

Ahhh, got it. Thanks!

[-]inklesspen14y20

I think the best definition of consciousness I've come across is Hofstadter's, which is something like "when you are thinking, you can think about the fact that you're thinking, and incorporate that into your conclusions. You can dive down the rabbit hole of meta-thinking as many times as you like." Even there, though, it's hard to tell if it's a verb, a noun, or something else.

If we want to talk about it in computing terms, you can look at the stored-program architecture we use today. Software is data, but it's also data that can direct the hard... (read more)

6PhilGoetz14y

I can write programs that can do that.

3thomblake14y

Philosophers love to make overly simplistic statements about what computers can't do, even when they're pro-tech. "Someday, we will have computers that can program themselves!" Meanwhile, a C program I wrote the other day wrote some js, and I did not feel like it was worth a Nobel Prize.

2ShardPhoenix14y

I think they mean a computer that can translate imprecise requirements into precise programs in the same way that a human can, not just code that outputs code. I do agree that philosophers can tend to underestimate what a computer can theoretically do/overestimate how wonderful and unique humans are, though.

0ShardPhoenix14y

I don't think anyone can yet write a program that can reflect on itself in quite the same way a human can.

5Perplexed14y

On the other hand, I don't know any humans who know themselves quite as precisely as does an optimizing compiler which has just compiled its own source to machine code.

3jacob_cannell14y

I like Hofstadter's works, but I think he over-focuses on recursion and meta-thinking. At a much more basic level, we use the word 'conscious' to describe the act of being aware of and thinking about something - I was conscious of X. Some drugs can silence that conscious state, and some (such as alcohol) can even induce a very interesting amnesiac state where you appear conscious but are not integrating long term memories, and can awake later to have no memory of large portions of the experience. Were you thus conscious? Clearly after awaking and forgetting, you are no longer conscious of the events forgotten. So perhaps our 'consciousness' is the set of all mindstuff we are conscious of, and thus it is clearly based on our memory (both long and short term). Even thinking about what you are thinking about is really thinking about what you were just thinking about, and thus involves short term memory. Memory is the key of consciousness, but it also involves some recursive depth - layering thoughts in succession. But ultimately 'consciousness' isn't very distinct from 'thinking'. It just has more mystical connotations.

[-]jacob_cannell14y20

I understand Computationalism as a direct consequence of digital physics and materialism: all of physics is computable and all that exists is equivalent to the execution of the universal algorithm of physics (even if an exact description of said algorithm is unknowable to us).

Thus strictly speaking everything that exists in the universe is computation, and everything is computable. Your option 1 seems to make a false distinction that something non-computational could exist - that consciousness could be something that is computable but is not itself comput... (read more)

1David_Allen14y

William Rapaport, in the paper PhilGoetz refers to, appears to exclude the idea that the universe is performing computation. He states: I would agree with you, that the universe is performing computation.

0PhilGoetz14y

I think you're basically saying "Option 2".

2David_Allen14y

The idea is that everything in the universe is a computation run by the universe. So yes, option 2 certainly. But functionalism describes a philosophy where the mind is formed by levels of abstraction. The substrate that performs the computation for any particular level is not important. This is option 1. So option 1 and 2 are not incompatible. They are context specific perspectives.

[-]DSimon14y10

(Warning: I expect that the following comment has at least one major error, since this topic is well outside my usual area of knowledge. Please read it as a request for edification, not as an attempt to push forward the envelope.)

Until we can detect or explain qualia in the wild, how can we make rational claims about its computability?

To make a simple analogy, suppose we have a machine which consists of a transparent box, a switch, and a speaker. Inside the box is a lightbulb and a light sensor. The switch controls the light, and the light sensor is hooked... (read more)

2jacob_cannell14y

I think you are onto the right idea with your analogy, but if you work through the implications, it should be clear that if qualia are truly not functionally important, than we shouldn't value them. I mean, to use your analogy - if we discover brains that lack the equivalent of the pointless internal light bulb, should we value them any different? If they are important, then it is highly likely our intelligent machines will also have them. I find it far more likely that qualia are a necessary consequence of the massively connected probablistic induction the brain uses, and our intelligent machines will have similar qualia. Evolution wouldn't have created the light bulb type structures - complex adaptations must pay for themselves.

1DSimon14y

I agree that qualia probably have fitness importance (or are the spandrel of something that does), but I'm not very sure that algorithms in general that implement probabilistic induction similar to our brain's are also likely to have qualia. Couldn't it plausibly be an implementation-specific effect, that would not necessarily be reproduced by a similar but non-identical reverse-engineered system?

2jacob_cannell14y

It is possible, but I don't find it plausible, partly because I understand qualia to be nearly unavoidable side effects of the whole general category of probabilistic induction engines like our brain, and I belive that practical AGI will necessarily use similar techniques. Qualia are related to word connotations and the subconscious associative web: everything that happens in such a cognitive engine, every thought, experience or neural stimulus, has a huge web of pseudo-random complex associations that impose small but measurably statistical influence across the whole system. The experience of perceiving one wavelength of light will have small but measurable differences on every cognitive measure, from mood to types of thoughts one may experience afterwards, and so on. Self-reflecting on how these associative traces 'feel' from the inside leads to qualia.

[-]novalis14y10

One of my favorite Philip Dick stories ("Gur Ryrpgevp Nag") is about the consciousness tape.

0Blueberry14y

That title is rot-13ed, in case that confused anyone.

[-]cata14y10

Whatever it is that's conscious, you can compute it and represent it in a static form. The simplest interpretation is that the output itself is conscious. So this leads to the conclusion that, if a Turing machine computes consciousness and summarizes its output in a static representation on a tape, the tape is conscious.

I don't see the contradiction here, although the sheer scale might be throwing off intuition. I'm perfectly willing to say that if you actually had such a tape, it would be conscious. Needless to say, the tape would be inconceivably huge if you really wanted to represent the astronomical amount of inputs and outputs that make up something we would call "conscious."

1jacob_cannell14y

Let's see - the retina absorbs roughly 10^7 bits/second, say you had to simulate just a day of possible input, that gives at least 2^1000000000 possible input/output combinations. There is 'possible in theory' and 'possible in any multiverse remotely similar to ours'. I tend to favor the 2nd usage of the world 'possible'.

0DanArmak14y

Since the algorithm can be compressed well (it fits into a human brain), and since that form of the algorithm takes its input a few bits at a time (and not a day's worth in a single go), it seems likely that a fully static representation can also be highly compressed and would not need to take the full 2^(10^7) bits. Especially so if you allow the algorithm to be slightly imprecise in its output.

0cata14y

Jacob was drastically oversimplifying, because the algorithm (assuming we restrict ourselves to responses to visual stimuli) does not convert one retinal image to some particular, constant output; a conscious being would never respond in the same way to the same image all the time. Instead, it converts one input brain state plus one retinal image to one output brain state, and brain states consist of a similarly enormous amount of information.

0DanArmak14y

Perhaps the difference between succeeding brain states, induced by visual input, isn't all that enormous.

[-][anonymous]14y00

Methinks the discussion suffers from lacking the distinction between consciousness-as-property: "subsystem C of system X is conscious-in-X" and consciousness-as-ontological-fact: "C exists in a conscious way" . Consciousness (of "C" in "X") is option-1-computed in the sense that it is a Platonic entity (as a property of platonically considered "subsystem C of system X"). It is option-2-computation in the sense that all such entities "C" could be said to perform various computations in "X" (and it is the ensemble of computations that the property detects). To draw moral conclusions ("C exists in a conscious way"), one needs to take X=reality.

[-]datadataeverywhere14y00

Sort of both. I think they reconcile more easily than you think.

Consciousness entities have behavior, including internal behavior (thoughts). Behavior obviously doesn't exist in stasis, which seems to be the point that you don't like about 1.

Consciousness is not the algorithm that generates that behavior, but rather that algorithm (or any equivalent algorithm) in action. It requires inputs, so that behavior can't just be read as a series of outputs and determined to be "the consciousness"; rather, consciousness is a property that entities operati... (read more)

[-]MartinB14y00

Option 4: there is no such thing as consciousness. Its all just an elaborate hoax our minds play on us. In reality we are 'just' state machines with complicated caches that give the appearance of a conscious mind acting. The illusion is good enough to fool us all, and works well enough for real life.

The more i learn off neuro science the more I get the impression that there is no real 'person' hidden in the body, just a shell, that runs lots of software over data storage, and seems to act somewhat consistent. So the closer you look, the less consciousness is left to see.

Please note that I still have trouble understanding how qualia work.

[-]NancyLebovitz14y130

My problem with that sort of explanation is that I don't see how there can be illusion without a consciousness present to be mistaken.

1atucker14y

Is an illusion something other than a system failing to correctly represent its otherwise accurate sensory input? Particularly if it happens in a specific and reproducible way, so that even though the information it gathers is accurate, said information is internally represented as being something different and inaccurate. Because I can write programs that have bugs that are pretty much that.

0NancyLebovitz14y

I'm not sure, but that's an interesting and possibly valid point.

8Tiiba14y

Something is making us talk about consciousness. Until we come up with a definition, we really have no business saying that this very real something is a fake version of some undefined ball of vagueness. It's like saying that trees are just big wooden things pretending to be snumbas. It's a lot better to just find out what makes us talk about consciousness and say, "Let's call it consciousness!"

1MartinB14y

Well yes. A good definition matters. Option 2 makes the least sense to me. It should be possible to have consciousness with any kind of computational device.

[-]David_Allen14y00

Consciousness is a roughly defined and (leaky) abstraction.

So this leads to the conclusion that, if a Turing machine computes consciousness and summarizes its output in a static representation on a tape, the tape is conscious.

It is the set of "stuff" that produces the consciousness abstraction that can be called conscious. In a Turing ma... (read more)

0PhilGoetz14y

As Nisan asked above: Is this Turing machine conscious if you don't run it?

1David_Allen14y

It seems that consciousness requires some type of thought, and that thought requires the system to self-modify. A static representation of the Turing machine then does not meet this requirement. So a Turing machine that is not running is not conscious. Is there another perspective to consider?

[-][anonymous]14y00

[-]Perplexed14y-20

If I were to attempt to characterize consciousness in computational terms, I would probably start with a diagram like that for the Mealy machine in this pdf. I would label the top box simply "computation" and the lower box "short term memory. I would speculate that consciousness has something to do with that feedback loop through short term memory. I might even go so far as to claim that the information flowing through short term memory constitutes the "stream of consciousness".

If this approach is taken, there are some consequen... (read more)

Moderation Log