Words as Mental Paintbrush Handles

Followup to:  Conditional Independence, and Naive Bayes

(We should be done with the mathy posts, I think, at least for now.  But forgive me if, ironically, I end up resorting to Rationality Quotes for a day or two.  I'm currently at the AGI-08 conference, which, as of the first session, is not nearly so bad as I feared.)

Suppose I tell you:  "It's the strangest thing:  The lamps in this hotel have triangular lightbulbs."

You may or may not have visualized it - if you haven't done it yet, do so now - what, in your mind's eye, does a "triangular lightbulb" look like?

In your mind's eye, did the glass have sharp edges, or smooth?

When the phrase "triangular lightbulb" first crossed my mind - no, the hotel doesn't have them - then as best as my introspection could determine, I first saw a pyramidal lightbulb with sharp edges, then (almost immediately) the edges were smoothed, and then my mind generated a loop of flourescent bulb in the shape of a smooth triangle as an alternative.

As far as I can tell, no deliberative/verbal thoughts were involved - just wordless reflex flinch away from the imaginary mental vision of sharp glass, which design problem was solved before I could even think in words.

Believe it or not, for some decades, there was a serious debate about whether people really had mental images in their mind - an actual picture of a chair somewhere - or if people just naively thought they had mental images (having been misled by "introspection", a very bad forbidden activity), while actually just having a little "chair" label, like a LISP token, active in their brain.

I am trying hard not to say anything like "How spectacularly silly," because there is always the hindsight effect to consider, but: how spectacularly silly.

This academic paradigm, I think, was mostly a deranged legacy of behaviorism, which denied the existence of thoughts in humans, and sought to explain all human phenomena as "reflex", including speech.  Behaviorism probably deserves its own post at some point, as it was a perversion of rationalism; but this is not that post.

"You call it 'silly'," you inquire, "but how do you know that your brain represents visual images?  Is it merely that you can close your eyes and see them?"

This question used to be harder to answer, back in the day of the controversy.  If you wanted to prove the existence of mental imagery "scientifically", rather than just by introspection, you had to infer the existence of mental imagery from experiments like, e.g.:  Show subjects two objects and ask them if one can be rotated into correspondence with the other.  The response time is linearly proportional to the angle of rotation required.  This is easy to explain if you are actually visualizing the image and continuously rotating it at a constant speed, but hard to explain if you are just checking propositional features of the image.

Today we can actually neuroimage the little pictures in the visual cortex.  So, yes, your brain really does represent a detailed image of what it sees or imagines.  See Stephen Kosslyn's Image and Brain: The Resolution of the Imagery Debate.

Part of the reason people get in trouble with words, is that they do not realize how much complexity lurks behind words.

Can you visualize a "green dog"?  Can you visualize a "cheese apple"?

"Apple" isn't just a sequence of two syllables or five letters.  That's a shadow.  That's the tip of the tiger's tail.

Words, or rather the concepts behind them, are paintbrushes - you can use them to draw images in your own mind.  Literally draw, if you employ concepts to make a picture in your visual cortex.  And by the use of shared labels, you can reach into someone else's mind, and grasp their paintbrushes to draw pictures in their minds - sketch a little green dog in their visual cortex.

But don't think that, because you send syllables through the air, or letters through the Internet, it is the syllables or the letters that draw pictures in the visual cortex.  That takes some complex instructions that wouldn't fit in the sequence of letters.  "Apple" is 5 bytes, and drawing a picture of an apple from scratch would take more data than that.

"Apple" is merely the tag attached to the true and wordless apple concept, which can paint a picture in your visual cortex, or collide with "cheese", or recognize an apple when you see one, or taste its archetype in apple pie, maybe even send out the motor behavior for eating an apple...

And it's not as simple as just calling up a picture from memory.  Or how would you be able to visualize combinations like a "triangular lightbulb" - imposing triangleness on lightbulbs, keeping the essence of both, even if you've never seen such a thing in your life?

Don't make the mistake the behaviorists made.  There's far more to speech than sound in air.  The labels are just pointers - "look in memory area 1387540".  Sooner or later, when you're handed a pointer, it comes time to dereference it, and actually look in memory area 1387540.

What does a word point to?

Comments

[Tom McCabe]

"What does a word point to?"

I'm not a neurologist, but I believe the classical answer is that words point to patterns in clusters of neurons. Everything filed under "apple" slightly rearranges neurons in the "apple" category, including pictures of apples, the taste of an apple, stories about apples, etc. Features which all apples have in common are amplified by the pattern-matching, while dissimilar features cancel each other out. Eventually, we get a mental pattern of the "archetypal" apple, describing the category in general, which can be drawn or described from scratch, without identifiable references to any specific apple. (If you haven't personally experienced this with apples, try ideas. Relying on one or two authors for your information will cause you to include some of their identifiable little details in your own writings.)

[Douglas Knight]

It's not the fault of the behaviorists. It's mentioned in:
Galton (1880), Statistics of mental imagery, Mind 5 p 301-18. (jstor, oup)

[FungibleFun]

You've been reading too much Hume.

[Yvain]

I had a professor, David Berman, who believed some people could image well and other people couldn't. He cited studies by Galton and James in which some people completely denied they had imaginative ability, and other people were near-perfect "eidetic" imagers. Then he suggested psychological theories denying imagination were mostly developed by those who could not themselves imagine. The only online work of his I can find on the subject is http://books.google.co.jp/books?id=fZXoM80K9qgC&pg=PA13&lpg=PA13&ots=Zs03EkNZ-B&sig=2eVzzMmK7WBQnblNx2KMVpUWBnk&hl=en#PPA4,M1 pages 4-14.

My favorite thought experiment of his: Imagine a tiger. Imagine it clearly and distinctly. Got it? Now, how many black stripes does it have? (Some people thought the question was ridiculous. One person responded "Seven. Now what?")

He never formally tested his theory because he was in philosophy instead of the sciences, which is a shame. Does anyone know of any modern psychology experiment that tests variations in imaging ability?

[Caledonian]

Words are just pointers. Concepts are the referents to which words make reference, and they are wordless things.

[Lee Corbin]

I knew a man who was chess champion of California who claimed that neither he nor his father could visualize anything. His father was a taxi driver in London, and long knew that the other drivers had mental maps, whereas he had to memorize lists of street names in just the right order. My friend, by the way, could also play blindfold chess just fine. He'd merely remember that "e7 is attacked on the diagonal from g5", presumably from long practice with the sequence e-f-g, 7-6-5. He lamented that his inability to visualize put an upper limit on his chess ability, and given how hard he worked, I believe it.

Galton was astonished that many scientists of his acquaintance did not visualize. Judging from Galton and other things I've read, the ability to visualize is more common today. TV, maybe. Please try to picture this without thinking of the old joke: "A man walked down the street, and turned into a drugstore". (The joke, of course, is that people do not become drugstores.)

Now did you happen to notice which side of the street he was walking on (left or right), and whether he was walking towards or away from you? The problem for those of us who use visualization for almost all our thinking, is that we must add irrelevant and often distracting information, which can be costly in math and science.

[Douglas Knight]

Yvain and Lee Corbin,
I tend to agree with Eric Schwitzgebel (which is how I found the Galton paper) that the difference in claims of visualizing ability is due to changing norms, not changing abilities. He's too quick to discount the possibility of real change, but professed inability of scientists in Galton's day is striking. I don't think that demographic has been adequately surveyed today, but I don't think they're as different from the general public as they were 50 or 100 years ago.

In particular, I don't think there was an early psychologist who couldn't visualize, as I wouldn't expect that to be so early as to have such a widespread affect on the scientists Galton knew. That is, I doubt a psychologist would have an impact on non-psychologists' self-assessment.

Also, I think people overrate the importance of visualization to navigation, blindfold chess, and other things where it seems intuitively important.

paper linked above:
"How Well Do We Know Our Own Conscious Experience? The Case of Visual Imagery" (2002), Journal of Consciousness Studies, 9, no. 5-6, 35-53.

[Ron Hardin]

What does a word point to? See an essay on words as labels in Stanley Cavell _The Claim of Reason_ p.175

In the background is always : what is this fantasy about? Meaning in this context the AI fantasy.

Actual robot fantasies begin around p.403

[Hendrik Boom]

other people were near-perfect "eidetic" imagers

I read a report in the Scientific American a few years ago in which they were doing experiments with rend0m-dot stereograms -- the kind of thing where if you just look at one image or the other you just see random dots, but if you look at one with one eye and the other with the other eye you see a square full of random dots floating above a background with random dots.

Some people could be shown one image one week, and the other the next week, each by itself, and suddenly get it. Evidently they had remembered the entire first-week image in memory to scompare with the second a week later.

I was impressed. This seemed eidetic enough for me.

[Caledonian]

Now did you happen to notice which side of the street he was walking on (left or right), and whether he was walking towards or away from you?

Your visualizations include such details? As the description didn't include such details, they're necessarily undefined - so why did you define them out of their uncertainty?

How many of you dream in concepts rather than images?

[Meta and Meta]

On behaviorism, I always liked Morgenbesser's alleged remark to Skinner: "Let me see if I understand your thesis. You think we shouldn’t anthropomorphize people?"

[Eliezer Yudkowsky]

Some people could be shown one image one week, and the other the next week, each by itself, and suddenly get it. Evidently they had remembered the entire first-week image in memory to compare with the second a week later.

I defy the data. That doesn't sound like it should be possible.

[Jack]

How many of you dream in concepts rather than images?

I've noticed this from time to time. It often seems that a dream will have a sense of urgency, or of being a child again, or of anything else, without any details from which this sense could be inferred. But it's not a flat 'all dreams' sort of thing; some dreams will be movie-like, others will be built out of pure feelings-that-something-is-happening.

[Tom McCabe]

"I defy the data. That doesn't sound like it should be possible."

I am also skeptical of this particular feat, but we know it's possible for the brain to record huge amounts of data in reasonably good detail in realtime. My theory is that everyone without a neurological impairment *stores* this data, but only a few rare people (savants and such) can access the data on request. To name a simple example, in one of my middle school classes, we had to read Dumas's The Count Of Monte Cristo. Very few people could read the book once or twice and recite the whole thing back in any kind of detail. Yet, when given a quotation from the book, most people were able to remember who said it, where they were at the time, who they were talking to, the surrounding context, etc.

[Caledonian]

I can produce vivid memories of images, sounds, textures, smells, etc. But my dreams are in streams of words - not the sounds of speech or the appearance of text, but the meanings alone. If I had a dream of the man walking down the street and turning into a drugstore, there would be no direction he was walking in, no perspective relative to me, and he wouldn't be turning left or right. He'd just be.

So you'd have a movie-like experience? Mine is more like reading a novel - only an invisible novel. While I have no body. Everywhere.

[Roko]

"What does a word point to?"

- unfortunately, I don't think you can say *in words* what the data structures of thought are. These structures and representations (and algorithms) need to be written down in maths.

and I don't think that the appropriate maths is probability theory or first order logic.

[that's a short explanation of why neat AI hasn't really worked in the last 50 years]

[Cyan]

My visualization of the man walking down the street included a number of irrelevant details: my view was from above eye level; the man was wearing a hat; he was walking towards me, but moving towards the right of me, not directly towards me; he was walking on the right side of the street (my right, his left), on a sidewalk; and he turned right (my right, his left) to go into the drugstore.

I believe my view from above was triggered by the word "down" in the description of the scenario, and the hat came from the word "old" -- the generic man-in-hat is also what I usually imagine when I read Nero Wolfe novels, which are set in the '30s through '70s.

I'm not claiming that these details a strongly fixed -- as soon as I consider alternatives to the specific details I related above, I get visualizations of those alternatives. I'm also not claiming that these visualizations are close to photographic in immediacy and sharpness -- for me, they have about the same clarity as visual memories, which is to say, not much.

[Brandon Reinhart]

Roko - "I don't think that" is not explanation.

[Alan Gunn]

I was astonished to learn years ago that some people read without "hearing" the words on the page; even today, though I know that this happens, it strikes me as odd. I even dislike reading the word "quay" because my first reaction is that it should rhyme with "way," and I know that it doesn't. Ditto with names that don't correspond to their spelling (Menzies, for instance--pronounced "mingiss" by Scots). And, perhaps relatedly, I have great difficulty visualizing anything, and never visualize anything clearly. I'm sure that there are genuine differences in visualizing ability--there are people who easily spell words backward by visualizing them and reading them off from right to left; I could no more do that than levitate. Richard Feynman somewhere described people at MIT learning to estimate the passage of time in different ways: some by counting (in their heads), others by picturing a moving tape with numbers on it.

[Roko]

@Brandon Reinhart:

True, true. I wish I had something more rigorous, but if I did then I would be writing a paper on it right now! All I have are some vauge intuitive ideas.

My intuition is that the key here is good knowledge representation systems. First Order Predicate Logic (FOPL) is good for something rather different than representing knowledge about the real world; it's good at representing statements about clean, abstract entities, namely about the truths of a particular formal system.

I think that my first intuition about the difference between statements of some formal language, and representations which would be useful to describe the real world is that FOPL statements don't come with a natural topology, and I think that one ought to have something like that. I think that a fairly key idea in the mathematics which underlies physics (calculus, etc) is the idea that we can approximate something. This is ultimately formalized by the metric on the real numbers, or on R^n. I seem to remember that those who invented calculus started by thinking about cutting things into very small slices, or about approximating a curve by small line segments.

I think that one needs some way of expressing the fact that two statements are conceptually close, and I don't think that something like mutual entropy or correlation of random variables does the trick. Why not? Well, think about the way a Taylor series approximates a curve (for example, the way it approximates the solution to a differential equation). You build up the representation of the (quite complex) real world thingy from a series of rather simple elements; the monomials (1, x, x^2, etc ) by taking a limit. Do we see probability theory allowing us to do this? Can I take a series of simple FOPL statements and approximate a messy, real world thing like a "fuzzy category" with them? I don't think so. Well, at least I've read some AI books and they all seem to limit themselves to encoding a real world concept as a particular FOPL statement, e.g.

Apple(X) <==> [ Green(X) or Red(X) ] and Edible(X) and Size(X, medium), etc.

This, I think, is the equivalent of thinking that the motion of a particle is ALWAYS some actual polynomial. In general, the motion of a particle is given by the solution to some system of PDEs or ODEs, which (by various theorems of analysis) (under nice circumstances) have a solution (on some open subset of R or R^n) which you can ultimately approximate by a polynomial.

I think that we're missing a whole chunk of theory about how you can iteratively build complex representations from simple ones, and how you can take limits of conceptual representations, and then how you can manipulate those limits.

[Latanius]

Are words really just pointers? If you want to refer to objects which you've visualized, they indeed are. But people even do some peculiar "arithmetic" with words, forming sentences, which has nothing to do with meanings.

For example, when I'm sleepy (half sleeping state), sometimes I notice that whole sentence structures are running through my head, without the words filled in, but I know where the sentences begin and end, and how they are connected. Even specific words show up time to time, but the whole stream has no sense at all. But if you don't visualize and use concepts... it just sounds right.

So I think words as stand-alone things (with their sound and syntactical role) also have an important role for connecting those things in a more abstract way, whose connections can't be inferred only by visualization. (Think of a linked list of pointers: the position of the pointers in the list can be as important as the referenced object itself.)

Roko:
FOPL is similar to the taxi driver who never visualizes anything. (It never dereferences the pointers.) I don't think the solution would be a much better symbolic system (although FOPL is not really designed for dereferencig), but to connect a visual cortex to the symbol manipulation system. So the similarity of two symbols could be checked by simply visualizing them.

[Ron Hardin]

Apple(X) <==> [ Green(X) or Red(X) ] and Edible(X) and Size(X, medium), etc.

The criteria for ordinary language making something count, or fit the case, are ordinary
language criteria, not mathematical criteria, of counting or fitting.

That is, ordinary language rules the operation of ordinary language, using the ordinary
meanings of count and fit, not the mathematical ones.

Ordinary concepts (nice red apple) are not less precise than mathematical concepts ; but
they give precision a certain shape.

The philosopher (not the mathematician!) wants to say that ordinary langauge
lacks something that mathematics has. The philosopher however is not curious about why he thinks this.

[michael vassar]

Field's Medalist insists that people don't visualize.
http://gowers.wordpress.com/2007/09/13/how-should-logarithms-be-taught/#more-5

The debate apparently goes on in educated circles to this day, and is compatible with high mathematical ability.

[michael vassar]

By the way Alan, it is possible to learn to read without hearing the words, and once you learn to do so you will probably find that you can read much faster that way. This won't help you to digest complex material faster, but will be useful when reading an article on a subject that you already know in hopes of finding some new and interesting information.

[Yelsgib]

Word as "pointer" implies the requirement for infinite storage - unless you say that "the word duck" is not a symbol referring to the symbol "duck" - or unless you claim that we can generate static memory spontaneously - or unless you believe that there is some special class or symbolized objects (like, you're -really- storing "duck" (content) somewhere but you are not storing "duck" (symbol) or "symbl duck" (symbol) somewhere).

Words, content, pointers, blah - it's all just computation until you can prove otherwise.

Do not presume structure.

[Alan Gunn]

Just a quick response to Michael Vassar: I am a very fast reader--just about the fastest I know. And I very much doubt that I could, at my advanced age, learn to read without hearing. Anyway, why would I want to? Among other things, I suspect that those who don't hear the words they read don't enjoy poetry as much as I do. What interests me about all this is that it seems to me to show that people's mental processes differ a lot more than we usually think--a topic that psychology doesn't seem to have paid a lot of attention to, and if the psychologists don't look into it, who will? (I don't know much about psychology, though; maybe my last point is wrong--hope so.)

[Caledonian]

I have a sneaking suspicion that most human talents come not from having mental resources that others do not, but by taking neural systems and tying them into different kinds of relationships with other modules.

I half suspect that the sections of my brain that are supposed to handle mathematics are devoted to processing language instead, for example.

Remarkable how many people confuse "what my mind does" with "what minds do". Perhaps this is related to the ubiquitous fallacy that mathematics has a Platonic existence beyond other phenomena and our conclusions are therefore universally binding.

[Roko]

@ Latanius: "FOPL is similar to the taxi driver who never visualizes anything. (It never dereferences the pointers.) I don't think the solution would be a much better symbolic system (although FOPL is not really designed for dereferencig), but to connect a visual cortex to the symbol manipulation system. So the similarity of two symbols could be checked by simply visualizing them."

Ok, and how do you visualize the concept "technology advances exponentially because technology feeds back positively on itself" or the concept "you can't define a word any way you like because the definitions you choose will bias your thinking in subtle ways", or even the concept of "status quo bias"

Not all concepts are amenable to useful analysis by the visual system. In fact, most of the concepts that are most useful (the general, abstract ones) are not amenable to this sort of analysis.

Futhermore, anyone who criticizes "symbolic" thinking would do well to read Marvin Minsky's book, "The Emotion Machine", in which he carefully explains why symbols are more useful then, say, connection strengths in a neural net, or (as you seem to be arguing for) pixel values in a bitmap image.

[Roko]

@Ron: Can you program your notion of "ordinary language criteria" into a machine? If so, you may well be using "mathematical" criteria to do so; this would be the case if you were using a programming language like c++ or prolog to write your program.

If you can't program your notion of "ordinary language criteria" into a machine, then I'm not really interested.

[Zubon]

I even dislike reading the word "quay" because my first reaction is that it should rhyme with "way," and I know that it doesn't.

Easily the most traumatic thing I have learned all day. I have been mentally pronouncing it "kway" for my entire life. I cannot recall having ever heard the word aloud. I reject your pronunciation and substitute my own! (The life of a reader is filled with many such traumas. Rendezvous? Epitome?)

[Ben Jones]

So, What does a word point to?

The same as anything else that enters our heads.

I look at a tiger. The actual tiger is not now in my brain. An image is represented and decoded in my brain. The relevant sensory cortex, cross-referencing with my long-term memory, flags up the concept 'tiger' and I'm done.

I look at the printed word 'tiger'. The resulting image is now represented and decoded in my brain. The relevant sensory cortex, cross-referencing with my long-term memory, flags up the concept 'tiger' and I'm done.

Same process when I hear "Tiger!", just via a different cortex. None of these things-that-enter-my-head is a tiger.

There is no hard, fast boundary between, say, a photorealistic image and a printed word. Neither of them is the referent. They're both representations. Hieroglyphics are the missing link, if you like. Granted, a word may have nothing objectively to do with what it refers to, but a smiley :) looks nothing like a face, and we have no trouble decoding that. Why? Because we share common knowledge of that meme/word before we look. Nothing magical or different about words. Just a slightly different mental process of decoding before the cross-referencing and concept-flagging.

Language may well be hard-coded into the human brain, but to some extent this is because 'language' is how we experince the world: via signifiers and representations (yes, even when looking). The 'language of abstract symbols' (i.e. words) isn't a separate magisterium!

Try for yourself. Taboo 'word' back to 'information that enters my mind through medium x'. Then see if changing x fundamentally changes what happens in your brain.

[Acheman]

Much of this discussion seems to be taking place within a somewhat naive conception of language and representation; in particular, it seems to me to be neglecting the insights that structuralism provided, overapplied though they may at some points have been. Contemporary linguistics, and much though sadly not all contemporary philosophy, recognise that language and thought operate both through the binding of words to images and then to 'things', and through distinguishing things and locating them within articulated systems. You can't have pointing, to give the example of a seemingly basic form of 'reference', without a system that distinguishes pointing-gestures from other motions of the hand, that defines the range of motions that count as pointing-gestures etc. I'm not going to get into a comment war about it, because there simply isn't space here, nor do I have the time, to have the sort of sustained argument that would be necessary to 'prove' this to someone used to thinking about language in another way - I can only recommmend more reading. Wittgenstein of course discusses this in the 'Philosophical Investigations', but - and I know I'm going to lose 90% of you when I say this, yet here you are anyway: one of the most lucid articulations of this double aspect of cognition occurs within psychoanalytic theory. I'm reading a very interesting book, 'Freud as Philosopher' by Richard Boothby, at the moment which I'm finding to be a particularly good treatment of this. To Eliezer in particular, I can't recommend this strongly enough - he provides a reading of Freud that's very different to the 'folk' one, and so far I particularly like his discussion of the relation of gestalt perception to the experience of mental imagery.
Since we're discussing our personal experiences, it came as a real surprise to me when I realised that some people couldn't hear the rhythm of words in their heads - having to count out the syllables of an iambic pentameter, for example. Another dimension I think it's interesting to consider here is imagining abstract concepts kinaesthetically. When I visualise the concept 'status quo bias', to use Roko's example, I imagine being subject to a force - a little like gravity, but stronger, maybe something like the feeling of holding a strong magnet away from something it's attracted to. I'm sure this is a heuristic that helps me to think more fluidly and intuitively about the concept.

[Alan Gunn]

Zubon: "(The life of a reader is filled with many such traumas. Rendezvous? Epitome?)"

I call your "epitome" and raise you a Yosemite (first encountered in Bugs Bunny comics; I thought for years it was "YOSE-mite"). Furrin words like rendezvous are OK, though.

[Joseph Knecht]

I personally saw the man walking away from me on the left side of the street, and my persective was just to the left of the curb on that side of the street and slightly higher than the man, who was a short distance from me. I saw him turn left into a drugstore for a split second, and then when I realized the joke briefly saw him morph into a drugstore on the sidewalk.

To the people who say that they visualized the scene but, for example, didn't see the person walking towards or away from you, or didn't see the man on one side of the street or the other: how can you visualize a man walking down a street and not at the same time have him be on one side of the street or the other (or in it), and was he not either facing you or facing away from you in the image? If not, was it that the man was vague and so didn't have features like a face that would indicate direction? I'm still puzzled how he could not have a position relative to the street in an image that includes both a person and a street.

[Caledonian]

Easy: it wasn't an image, and we didn't visualize it.

Thus: no position, no perspective, no direction, no facing. No features, either. Just a man walking down a street and turning into a drugstore.

[Joseph Knecht]

Caledonian: you said "Your visualizations include such details? As the description didn't include such details, they're necessarily undefined - so why did you define them out of their uncertainty?"

I understood from your statement that you expressed surprise that the reported visualization contained such details as "which side of the street the person is walking down". This implied to me that you believe it is possible to visualize a man walking down a street, but not be either walking down the left or right side or in the street itself, etc.

[Caledonian]

No, not really. I was just surprised that people attempted visual-style representations, complete with details, without having sufficient data for even an approximation.

If I'm specifically asked to visualize a triangle, I can do so - and then tell you how it's oriented, what color it is, and what color the background is. None of those things are really implied, and they're not necessary either. But if I just hear a triangle mentioned, I perceive none of those things. The thought that someone *would* is strange to me.

[Cyan]

It's not like I have a choice, dude.

[Latanius]

@Roko: The visual cortex isn't the only one thing we use. Other parts of the brain probably "cache" some of the insights gained by visualizing things, or trying / imagining movements etc., also common sentences, so we can use these areas for other things we've never seen before. These cached things are our concepts, I think.

You're right, I won't visualize every part of the thought "technology advances exponentially because technology feeds back positively on itself". But I've seen a lot of exponential functions in math classes, plotted them on screen, and noticed that they can grow very big. Now I use this concept for understanding this sentence. It would be hard to explain this to a five year old, or to somebody who has never seen exponential functions: you can't visualize so many things at once, without using any cache mechanisms. (That's why inferential distances are so long in reality, I think.)

With only language and the cached thoughts (grammar / logic and rules in a symbolic system) we can get surprisingly far, but not far enough. (For us, even logic is a cached thought from the visual cortex, for it describes the connections of distinct things. This is a special feature of vision: try to imagine two songs at the same time...)

[Roko]

@latanius:

sure. But what I'm saying is that you need some mathematical machinery to manage all of these interactions between pictures, sentences, sounds, etc. There's no shortage of narrow AIs which make use of one aspect of what you've said, e.g. face recognition "AIs", text-matching AIs (google), etc. But none of them can pass the Turing test.

Furthermore, I think that there are abstract concepts (like "bias") which are not well represented by anything we have at the moment. I think that we need to forge new abstract representations, and that those representations will be symbolic, they will look more like FOPL than like a neural net or a bitmap.