My Ghost Story

[JayUtah]

I lean toward thinking that she did not realize the muddled nature of her thinking until it was pointed out here, and now she is having difficulty accepting it.

Indeed it's difficult (and usually pointless) to wonder why people are projecting, backpedaling, or trying to save face. It's usually sufficient to accept that, for whatever reason, those exercises are taking place and that they result in untenable arguments. We can then address the arguments on their face.

Compare Song’s very direct claim that computers will never be able to model consciousness with the second link about Koch in post 460 below; the two are in direct opposition.

The invocation of computer science here is also largely a red herring.

What's important to realize from the contradiction you've identified is that Jodie's body of science, from which she plans to depart into speculation, is not the solid rock from which speculation can launch upon a productive course. Her speculation is not "based on science" but instead based on others' unproven speculation. As eminent as these scholars may be, they have reached opposing opinions on important points because they have proposed incongruent conceptual models and adopted different assumptions within them.

The postulate of consciousness as finite closed system, albeit requiring quantum dynamics to describe (cf. Tegmark), would seem to sanction the potential of an automaton to incorporate that description. Song's denial is moot. A description of the behavior of a system is not necessarily a description of its operation. A model of a system, especially through sequential automata, rarely captures the operation; it focuses instead on behavior. An automated model would be no more an operative consciousness than the automaton from the film Hugo is conscious to the meaning of the picture it draws. Modeling something and being something are two qualitatively different things.

Article on Tegmark and perceptronium

...

Tegmark, like Koch, does not posit consciousness as something separate from matter, [...] According to Tegmark, consciousness is not a free floating thing wafting between dimensions; rather, it can be explained just like physical matter.

There is such an incongruence of concepts in Jodie's speculative formulation that it's difficult to know where to start.

First, I guess, "artificial intelligence" isn't this. Artificial intelligence is a set of reasonably well understood methods for creating automata that mimic to some useful degree the behavior of an intelligent being. It is no more equivalent to machine consciousness than organic intelligence is equivalent to organic consciousness. Specifically, artificial intelligence employs techniques pertinent to the vocabulary of large-scale Turing-style automata, that usually have no structural or procedural analogue in the human brain. For that reason, machine consciousness isn't really a thing. Among people who work with "intelligent" machines, the notion of whether they can have or model consciousness is almost entirely moot. It's relegated to science fiction.

We already discussed modeling versus recreating.

Tegmark's 2015 paper describes consciousness hypothetically as a finite, probabilistic state of matter. He specifically declines to prove the model. He merely states, based on a set of initial conjectures, what could and could not follow from them according to our current formulations of statistical physics. Notions of independence and autonomy bubble up through that formulation. He does not define the boundary of the system in connection with any of these. He merely discusses what could differentiate conscious matter from nonconscious matter without offering concrete examples.

His finding regarding independence allows that a body of conscious matter must exhibit some degree of coherence and a degree of independence from its environment, but does not dictate it must be entirely independent. In fact, he argues that conscious matter cannot be perfectly independent. One might be tempted to map this in concrete terms to hypothetical concepts such as universal consciousness or transcendent consciousness. But that would greatly misrepresent Tegmark's intent. The author insinuates that dependence (or interdependence) in conscious matter is merely "environmental" without elucidation. There is nothing in the abstract principle of consciousness being affected by environmental factors that allows for or directs that those factors must be any sort of supernatural.

Overall there is nothing in Tegmark's paper that would justify specific conjectures such as universal consciousness among living beings, consciousness being affected via means we don't presently know, conscious influence through some Hilbert-as-the-universe mumbo-jumbo (or any other multidimensional model of our universe, which he doesn't even touch), or temporal mechanics suggestive of prophecy. None of what Jodie proposes in her interpretation of her dream has the slightest support (or even, for that matter, mention) in Tegmark.

 

[Garrette]

Indeed it's difficult (and usually pointless) to wonder why people are projecting, backpedaling, or trying to save face. It's usually sufficient to accept that, for whatever reason, those exercises are taking place and that they result in untenable arguments. We can then address the arguments on their face.



The invocation of computer science here is also largely a red herring.

What's important to realize from the contradiction you've identified is that Jodie's body of science, from which she plans to depart into speculation, is not the solid rock from which speculation can launch upon a productive course. Her speculation is not "based on science" but instead based on others' unproven speculation. As eminent as these scholars may be, they have reached opposing opinions on important points because they have proposed incongruent conceptual models and adopted different assumptions within them.

The postulate of consciousness as finite closed system, albeit requiring quantum dynamics to describe (cf. Tegmark), would seem to sanction the potential of an automaton to incorporate that description. Song's denial is moot. A description of the behavior of a system is not necessarily a description of its operation. A model of a system, especially through sequential automata, rarely captures the operation; it focuses instead on behavior. An automated model would be no more an operative consciousness than the automaton from the film Hugo is conscious to the meaning of the picture it draws. Modeling something and being something are two qualitatively different things.



There is such an incongruence of concepts in Jodie's speculative formulation that it's difficult to know where to start.

First, I guess, "artificial intelligence" isn't this. Artificial intelligence is a set of reasonably well understood methods for creating automata that mimic to some useful degree the behavior of an intelligent being. It is no more equivalent to machine consciousness than organic intelligence is equivalent to organic consciousness. Specifically, artificial intelligence employs techniques pertinent to the vocabulary of large-scale Turing-style automata, that usually have no structural or procedural analogue in the human brain. For that reason, machine consciousness isn't really a thing. Among people who work with "intelligent" machines, the notion of whether they can have or model consciousness is almost entirely moot. It's relegated to science fiction.

We already discussed modeling versus recreating.

Tegmark's 2015 paper describes consciousness hypothetically as a finite, probabilistic state of matter. He specifically declines to prove the model. He merely states, based on a set of initial conjectures, what could and could not follow from them according to our current formulations of statistical physics. Notions of independence and autonomy bubble up through that formulation. He does not define the boundary of the system in connection with any of these. He merely discusses what could differentiate conscious matter from nonconscious matter without offering concrete examples.

His finding regarding independence allows that a body of conscious matter must exhibit some degree of coherence and a degree of independence from its environment, but does not dictate it must be entirely independent. In fact, he argues that conscious matter cannot be perfectly independent. One might be tempted to map this in concrete terms to hypothetical concepts such as universal consciousness or transcendent consciousness. But that would greatly misrepresent Tegmark's intent. The author insinuates that dependence (or interdependence) in conscious matter is merely "environmental" without elucidation. There is nothing in the abstract principle of consciousness being affected by environmental factors that allows for or directs that those factors must be any sort of supernatural.
Overall there is nothing in Tegmark's paper that would justify specific conjectures such as universal consciousness among living beings, consciousness being affected via means we don't presently know, conscious influence through some Hilbert-as-the-universe mumbo-jumbo (or any other multidimensional model of our universe, which he doesn't even touch), or temporal mechanics suggestive of prophecy. None of what Jodie proposes in her interpretation of her dream has the slightest support (or even, for that matter, mention) in Tegmark.

Thank you. You clearly understand these things to a greater degree than I do, but so far I'm able to follow. The highlighted bit is what I was trying to get at, too. A superficial reading of a non-scientist's restatement of Tegmark or Koch or even Song can easily lead a layman to think otherwise, i.e., to think that consciousness is separate from matter and can traverse dimensions (not trivially including time). But such a reading betrays itself and indicates that the layman's understanding really is one of headlines and chapter headings perhaps sprinkled with catchy new-age internet memes. They are seductive.

 

[JayUtah]

The highlighted bit is what I was trying to get at, too.

Yes, it bears additional underscoring.

Tegmark's original paper does all its work in Hilbert space. That's not inappropriate, because what can be proven to work for Hilbert space works in all subspaces, by definition, including Euclidean space and the conceptual vector spaces Tegmark invokes to model the various properties he postulates for consciousness. Proofs in Hilbert space are messy, but powerful. But that's just an artifact of the author's approach.

The upshot is that the "many dimensions" that create the nastiness of those proofs have absolutely zero to do with "dimensions" in Jodie's speculation. It's just the same word used in different connotations.

 

[Garrette]

uh.....I completely understand what Hilbert space is. And those conceptual vector spaces, too. No, really. I do.

stop laughing

 

[JayUtah]

uh.....I completely understand what Hilbert space is. And those conceptual vector spaces, too. No, really. I do.

stop laughing

No one laughs when talking about vector calculus. But the point is that it's easy for new-age claimants to point to some impressive-looking agglomeration of complicated mathematics that they themselves do not really understand, and insinuate that something in it allows their beliefs to be a rational possibility. The gambit is that their critics either don't understand it themselves and therefore cannot rebut, or that they cannot explain the rebuttal in unequivocal terms.

You had the right answer the first time. Vague references to complicated math do not automatically vindicate every fringe theory whose proponent makes the reference.

 

[Garrette]

Agreed.

 

[Donn]

Thank you both JayUtah and Garrette. You raise the tone and keep failing courage alive.

 

[Cosmic Yak]

Nah, they're just knee-jerk, emotive responses.

 

[Cosmic Yak]

Jodie insinuates she can take a red shade of science and a blue shade of speculation and mix them into a lovely violet shade of acceptable sciency-speculation. It's more accurate to say you can't take the dough of speculation and the trumpet of science and smoosh them together, expecting a cookie that plays mariachi music.

Seriously, though, this was brilliant!

 

[Donn]

Seriously, though, this was brilliant!

Yes, quite! Distilled wisdom and wit.

 

[Daylightstar]

No one laughs when talking about vector calculus. But the point is that it's easy for new-age claimants to point to some impressive-looking agglomeration of complicated mathematics that they themselves do not really understand, and insinuate that something in it allows their beliefs to be a rational possibility. The gambit is that their critics either don't understand it themselves and therefore cannot rebut, or that they cannot explain the rebuttal in unequivocal terms.

You had the right answer the first time. Vague references to complicated math do not automatically vindicate every fringe theory whose proponent makes the reference.

Very, very well put!

From which the below bears repeating:
"... new-age claimants to point to some impressive-looking agglomeration of complicated mathematics that they themselves do not really understand, and insinuate that something in it allows their beliefs to be a rational possibility. ..."

 

[JayUtah]

uh.....I completely understand what Hilbert space is. And those conceptual vector spaces, too. No, really. I do.

I wanted to expand upon this a bit more. Not that you'll suddenly understand Tegmark. But you may understand why he did it.

This started out in my head as an elementary essay on linear algebra. But as that tends to make people drowsy, I'll just try to provide clarity and a few examples. We tend to teach linear algebra either by introducing Euclidean space as an intuitive example of a vector space (which then limits abstract use) or by relentless use of notation. I've never been a fan of introducing mathematical concepts by first defining the notation and then expecting people to grok the concepts as you scribble the notation out. It's like teaching someone an entirely new language in the morning and then expecting them to appreciate poetry written in it by afternoon.

Euclidean space is simply our three-dimensional world. We measure it according to three orthogonal axes corresponding to height, width, and depth. The component measurements along each of those axes constitute a vector in the space. Vectors are to a vector space what plain old numbers are to the number line. (The number line is also a vector space, but we haven't caffeinated enough for that.)

Numbers on a line have operations (addition, multiplication, etc.) and an algebra (well-behaved symbolic manipulation of expressions composed of those operators). Those give rise to functions (uniform mappings of some values on the line to other values on the line), and then a calculus (well-behaved symbolic manipulation of the functions). All this creates a powerful set of tools for describing how quantities behave, that we represent with those numbers.

Ditto vectors. They have operations, an algebra, and a calculus. By describing our world as a Euclidean vector space, we can use those tools to reason about the geometry of our world -- the actual real-world geometry, such as how much gravel do I need to buy to cover my driveway. But that's only one example of a useful vector space. For most of us it's by far the most useful one. But we can represent with a vector quite a number of things where a data element is composed of two or more related values. We consider these "conceptual" vector spaces because they don't always work in intuitive Euclidean ways.

In control system design, we often consider a 2-dimensional vector space where the first coordinate is the "error," or how much the measured variable differs from the desired value. The second coordinate is how fast it's changing. If you then visualize it as a 2-D coordinate system, error is the x-axis and error rate is the y-axis. Every state of the system can be plotted as a point on that graph. The system is stable if its current state is close enough to the origin to make you happy. We then use the mathematics of vectors to reason about the effects of inputs to the system, and to investigate its behavior over time as paths through that 2-D space.

Quantum mechanics is one of the sciences whose interesting values are best expressed as vectors. Not literal Euclidean vectors, of course. Not spatial vectors. Purely conceptual vectors. We speak of the multiverse having 11 dimensions. That doesn't mean literally 11 mutually orthogonal spatial dimensions, so put away your copy of Flatland. Like our control system example above, the coordinates are related and congruent but not identical in meaning. A variable and its time derivative are certainly related, but not identical concepts.

So most of the investigation and reasoning with respect to quantum mechanics is done in various conceptual vector spaces. Then we can use vector and matrix operations to write expressions that describe how we believe the system behaves that is described by vectors in the space. We do it symbolically (algebraically).

I threw out the term "orthogonal" above. In Euclidean terms, it simply means right angles. Height is completely at right angles to width and depth. I can change my height without changing either one of the other two. An example of non-orthogonality would be trying to adjust the shower temperature when you have only the two hot and cold knobs, and your vector space is composed of temperature and flow rate. You turn up the hot knob, and yes you get hotter water but the flow rate also increases. Conceptually, "the hot knob" corresponds to only one of the coordinates in Shower space (temperature). In order to change only of the coordinates in Shower space (either temperature or quantity) while leaving the other alone, we have to manipulate both knobs. That makes it nonorthogonal.

Lots of vector spaces we use productively are not orthogonal. Math to determine which coordinates are relatively independent and which aren't help us understand how a system behaves. Specifically, Tegmark uses one of those to determine what a vector-described quantum system would have to look like in order to be considered an independent system (i.e., one of his criteria for consciousness).

So who is this guy Hilbert? The vector spaces we have discussed so far have finite dimensionality. The real numbers are a (now boring) 1-dimensional space. Euclidean space has 3 dimensions, and we find the 2-dimensional restricted subspace of it to be highly useful in practice too. Our conventional control system has 2 dimensions. Because there is a limited number of dimensions, most of the operations we do on them exist in closed form: equations where you plug in the values of the vectors and a determined answer emerges via simple arithmetic. At worst, you have a system where you have to repeat the same operation on each coordinate.

Hilbert space is the generalization of vector spaces to infinite dimensions, with certain restrictions on the operations you can do in it. It has the very helpful property in that all conceivable vector spaces that apply to a given problem (let's say, oh, quantum mechanics) are contained within the Hilbert space and obey all the same rules regardless of dimensionality. What can be proven for a given Hilbert space must hold for all its subspaces. It has the very annoying property of being infinitely composed and therefore not susceptible to simple closed forms or finite iteration.

Once you've been beaten over the head enough by various branches of mathematics, you realize that once you start talking about infinite compositions, you're into calculus. Among other things, calculus lets us convert repeated sequences of discrete operations (e.g., do this for the x-coordinate, then the y-coordinate, then the z-coordinate,...) into one transformed algebraic form. Therefore techniques in vector math that are simple with a finite number of dimensions become more complicated as you introduce the calculus, which requires eigenvectors, eigenvalues, and increasingly more esoteric mathematical credit cards by which you expend simplicity in order to gain power.

But that's why Tegmark goes there. He doesn't solve just one incarnation of the quantum mechanics problem in terms of what he wants to know; he solves all possible incarnations of it. In the math world (and also the physics world), solving a problem in its most generalized state is preferred. It means a solution that has maximum value, but you have to know quite a lot about the underlying math to see exactly what it proves and how.

 

[Garrette]

Thanks. That did help, though probably not as much as you'd hoped for the effort. Remind me some time to tell you about my advanced math placement experience as an undergrad.

 

[JayUtah]

Very, very well put!

From which the below bears repeating:
"... new-age claimants to point to some impressive-looking agglomeration of complicated mathematics that they themselves do not really understand, and insinuate that something in it allows their beliefs to be a rational possibility. ..."

Thanks. There's an insidious rhetorical dialogue in a lot of new-age argumentation, some of which has occurred here.

Imagine one of your employees comes to you and asks for a raise.

"Why do you think you deserve a raise?" you ask.

"Because this teddy bear," says the employee. And he places a stuffed bear on your desk triumphantly. You think for a moment, trying to discover the rational connection between the teddy bear and the request. But of course there isn't any. And you say so.

"Sorry, nothing about that teddy bear affirms that you're entitled to a raise."

Unsatisfied, the employee retorts, "But nothing about the bear prevents me from having a raise, right?" Irrelevancy is just that. It neither affirms nor denies. But the amphibolistic about-face has the rhetorical effect of shifting the presumptions of relevancy in a way that seems convincing without being at all probative. It's a clevery-disguised complex question.

What the employee really says is: "Here is a bear. It is relevant to the question. Specifically, it affirms my right to a raise." If you mistakenly rebut only the conclusion, then the premise still hangs there -- unproven but tacitly accepted ("The bear is relevant, but you are not entitled to a raise."). The proper answer is, "The bear is irrelevant." The justification for a raise must begin anew, with an entirely different set of premises.

The proper answer here is that most of the speculative works to which Jodie has alluded are simply irrelevant to her proposition. They neither support nor deny her claim, because that's what it means to be irrelevant. But by her entry of them into evidence, she has floated the axiom that they are relevant. And if they're taken as relevant, then their lack of denial seems somewhat probative.

What of those that are relevant, even perhaps marginally so?

Let's change the thought experiment. Your employee asks for a raise as before, but instead justifies it with, "I think my work here merits a raise." There you go -- that's a rational connection.

But with a tip of the hat to Garrette's observation, let's say this summit followed a lengthy discussion between the employee and his wife, coworkers, and other members of the industry. And let's say that conversation built the employee's esteem to the point of submitting it for more impartial judgment. Emboldened by praise from his peers, he stands before you now with a syllogistically valid argument and a belief about the data.

"Well, let's talk about that," you say. You pull up his attendance records and note that he arrives late and leaves early frequently, and takes many days off at short notice. You see that his customer satisfaction rating is about average, and that his work products are susceptible to a high rate of required rework. "Based on what I'm seeing here, you don't merit a raise."

Some things just don't mean what we think they mean, and sometimes we just aren't as adept according to some relevant set of metrics as we believe in our hearts subjectively. Things that appear relevant, such as coworker praise or family encouragement, aren't necessarily the standards that really matter. The employee can either respond emotionally, or he can try to understand what the standards are and why.

 

[Donn]

Non sequiturs, at base. A shower of shinola, thinly coated in ersatz sci-fi, forms the backdrop to a plea from hidden motives for us to board the fancy train and freewheel.

 

[Jodie]

Your critics are right. You can't decide whether your claim is science or speculation, so you vigorously tap-dance between the two.

I think I'm pretty clear where the science ends and my speculation begins. If you didn't bother to read any of the articles I linked or read anything by the authors I listed I can see why you would be confused on this point.

 

[Jodie]

No part of that sentence makes sense.

Nothing you are claiming happened ever happened.

You are making up asurdities in order to explain away impossibilities.

If you don't understand the science behind the statement then it wouldn't make sense to you. Go back and read what Tegmark and Song had to say about consciousness and what indications we have here in our 4D world that multiple dimensions exist. After that, I think you'll understand where my speculation starts.

You are welcome to believe whatever you want to believe about my dream but it did happen. I have no reason to make up anything. I like my idea but whether or not I have it right remains to be seen.

 

[Jodie]

Your idea comes from your interpretation of what you think the research in various fields says, as filtered through the preconceived result you're looking for. I'm afraid I can't see anything in your posts to suggest you've accurately understood current thinking in psychology, physics or AI, never mind managed to somehow link them all.

Do you mind clarifying what areas of psychology, AI research, and physics I've misinterpreted ? You don't have to go into great detail but a few statements with links that I can use to do follow up reading would be appreciated.

 

[JayUtah]

I think I'm pretty clear where the science ends and my speculation begins.

You claim your speculation is "based on" these items. It is not, as has been amply explained.

If you didn't bother to read any of the articles I linked or read anything by the authors I listed I can see why you would be confused on this point.

There is no confusion. Your sources simply don't support your speculation.

 

[JayUtah]

Do you mind clarifying what areas of psychology, AI research, and physics I've misinterpreted ? You don't have to go into great detail...

But I have. Please address the several lengthy articles that have been posted since you last arrived.