Chomsky / Surfaces and Essences

This post compares Noam Chomsky's Linguistics to the ideas put forward in "Surfaces and Essences" - S&E- by Hofstadter and Salter. A good introduction to Chomsky's theories can be obtained in:

"Noam Chomsky on Linguistics"

For anyone new to Chomsky, I suggest that you read S&E first. While Chomsky is confident that analogies (the core of S&E) are irrelevant to the study of linguistics, anyone familiar with S&E will find themselves shouting at Chomsky's video.

THE INNATE CAPACITY OF LANGUAGE

It seems to me that Chomsky's theory harks back to an era when theories of computation were in vogue, especially at Chomsky's home base, MIT. His linguistic theory (Universal Grammar) is closely analogous (oops) to fledgling formalism behind the new computer languages being developed in the 1950's. His core assumptions seem to be:

1. Language is an innate capacity of humans;
2. Language is computational in its essence. It is a computational process carried out by the brain.
3. Linguistics, as conceived by Chomsky, is a search for "the" fundamental "algorithm" behind language. The video mentioned above gives a taste of the formalism invented by Chomsky to "parse" speech in a way that supposedly doesn't depend on the language being used.

The "innate" capacity is distinguished from 

1. The perceptual systems that deliver stimulus to the brain - sounds or sights - that are the raw material to for language input and 
2. The physical mechanism of speech
3. Any particular language. Chomsky maintains that a baby starts to distinguish the unique features of his parent's language very early - possibly in the womb, but that any baby can learn any particular language.

Chomsky has had over 1/2 a century to flesh out these ideas, to defend them and turn them into something useful - a basis for a "Scientific" theory. The above presentation leads me to believe that he hit a dead end.years ago.

One positive aspect of Chomsky's approach is that he freely admits that there is a lot more going on in the mind besides language. One misses this kind of humility in S&E.

EVOLUTION

Whenever anybody invokes evolution to support or explain any aspect of his theory, red flags should go up. Chomsky is no exception. He prefaces his ideas on evolution by dismissing the public's "pop" idea of how evolution works, but it's hard to see how his own ideas differ. He still seems to be wedded to the idea that any aspect of an organism (phenotype) must be explained in terms of an adaptation beneficial for survival. This backs him into a corner where he feels that he must account for his (hypothetical) unique language capability in terms of what amounts to "pop" evolutionary theory. By his own (apparently unnecessary assumption), this capability must have popped into existence suddenly, "out of the blue" and not through the assembly of gradually-acquired mutations (such as increased area of the cerebral cortex).

One of Chomsky's key claims is the insistence that language is unique to humans, without parallel or precedent in the animal kingdom. Another related (but strangely irrelevant) claim is that language is the necessary and sufficient evidence of abstract thought. Thus, for example, he assumes that archaeological evidence of symbol making are evidence of language. This forces him into making silly statements such as his claim that dogs can't have ideas or make generalizations. This is only true if you define ideas and generalizations as language, which is preposterous.

He claims (needlessly) that the innate capability of language appeared recently and suddenly in evolutionary history, possibly by a single mutation. Obviously, language would be not much use to the single person who had this mutation. Of course, a mutation will quickly spread through a population, even if its "survival advantage" is neutral. If you think carefully about this, the argument that this capability appeared "suddenly" becomes meaningless. Any mutation can spread through a population after appearing only once (red hair for example). "Selection" will operate against the mutation only if it's harmful. Chomsky introduces evolution to "sound scientific". It's irrelevant to his theory.

The frog's eye generalizes what is perceived and sends a signal that amounts to "food" or "danger" to the frog's brain. That's analogy making. S&E is founded on analogy and has no need for anything to pop into existence suddenly with humans. Human analogy making can be "explained" by gradual improvement in the ability to make good analogies (the "Essence" in "Surfaces and Essences), which could very plausibly be associated with growth of the amount of grey matter devoted to analogy making throughout the brain.  Chomsky has no idea what the language "capability" is working with (certainly not analogy), so his theory is built on a huge, sudden, mysterious mutation. This isn't much different from a "Gift from God".

ACQUISITION OF LANGUAGE

Chomsky has been roundly criticized by his failure to account for the way real humans acquire language, yet he admits that any linguistic theory must fully account for how linguistic ability develops in humans. Listening to Chomsky, one gets the idea that it's been a long time since he's had a conversation with a two-year-old. S&E claims that children learn language by analogy - copying speech patterns and learning the situations where these speech patterns are appropriate, then (automatically and somewhat mysteriously) developing vast stores of analogy that allow them to use the same or similar noises to apply to "similar" situations. This makes a lot of sense and agrees with what we see in actual children. Chomsky's theories about how children learn syntax, especially how they form sentences into words, seems like a wishful fable in comparison.

REDUCTIONISM

Reductionism is still at the basis of so much "scientific" thinking that it's hard to spot in Chomsky's thinking. However, it's clear that his project is to reduce language to something simple and almost "atomic". In fact, he uses the example of chemistry, which was initially a project to simplify the principles and identify the underlying "rules". The idea that all phenomena can be "reduced" to effects that emerge from the actions of simpler, more fundamental systems has been discredited in the 21st century. We no longer see the need to "boil down" brain capabilities to "computations" in a meat computer between our ears. They can be studied on their own. There is nothing stopping us for explaining or describing these capabilities in terms of more fundamental processes in the brain, but we need not assume that such explanation is possible "in principle".

Chomsky is clearly in awe of the "hard sciences" and hopes that linguistics will evolve in that direction. However, he admits that that is not actually happening. In fact, the attempt is falling apart, partly due to its premature efforts to find formal "laws" that simplify the staggering quantity of data it tries to "explain". Linguistics is more complicated than the phenomenon it tries to explain/

One suspects that perceptions and analogies are the very "atoms" Chomsky is looking for to form a foundation for a theory of linguistics. But it's too late. He's explicitly rejected this option and clings to his theory that, by his own admission, is built on mysterious interfaces between language and the real world.

APPLICATIONS

To my mind, the biggest difference between S&E and Chomsky's theories is in the area of application. One cannot read S&E without gaining a whole new insight on language and the workings of the human mind. Applications abound, from teaching children in Kindergarten to analyzing the bad ideas that cause wars. There is no hint of application if Chomsky's linguistics in the video mentioned above, nor have I ever heard mention of a practical application. Failure to apply in the case of children has been noted.

I'm also reminded of a conversation I had with a young lady studying linguistics in University. She was totally unable to explain what her subject was about. This echos the situation with students studying philosophy or string theory. These are all subjects "about" themselves.

CONCLUSION

There is some value in going over Chomsky's breakdown of the problems involved to see how S&E solves them. For example, S&E provides a seamless transition between perception and language. It provides a similarly seamless transition between non-human "thinking" and human thinking. Chomsky is mystified by the problem of "free will", which boils down to extending our idea of thinking to cover action. This seems to pose no problem in S&E but is a huge problem for Chomsky, who regards the idea that thought and action are the same kind of thing as a throwback to behaviorism.  In fact, Chomsky himself has a very hard problem with categories and analogies even though he can't seem to talk for five minutes without supporting his own projects by means of analogy.

S&E would seem to sweep Chomsky linguistics aside. Whether or not S&E is "true" in detail, it is clearly a more useful and productive way to think about language. Chomsky's linguistics is already under attack for reasons that have nothing to do with S&E (it simply doesn't work), but S&E deals many of the mysteries and unsolved problems of S&E. To my mind, there is not much left of interest.

It should be remembered that both S&E and linguistics leave aside the realm of human thought that is not connected to language. Generally speaking, all the capacities associated with the "self" are presumably inbuilt capacities of the brain which remain intact even if speech capabilities are lost (for example, by a brain lesion in the left brain). While study of language teaches a lot about the human mind, it's not the whole story.

Chomsky / Surfaces and Essences

This post compares Noam Chomsky's Linguistics to the ideas put forward in "Surfaces and Essences" - S&E- by Hofstadter and Salter. A good introduction to Chomsky's theories can be obtained in:

"Noam Chomsky on Linguistics"

For anyone new to Chomsky, I suggest that you read S&E first. While Chomsky is confident that analogies (the core of S&E) are irrelevant to the study of linguistics, anyone familiar with S&E will find themselves shouting at Chomsky's video.

THE INNATE CAPACITY OF LANGUAGE

It seems to me that Chomsky's theory harks back to an era when theories of computation were in vogue, especially at Chomsky's home base, MIT. His linguistic theory (Universal Grammar) is closely analogous (oops) to fledgling formalism behind the new computer languages being developed in the 1950's. His core assumptions seem to be:

1. Language is an innate capacity of humans;
2. Language is computational in its essence. It is a computational process carried out by the brain.
3. Linguistics, as conceived by Chomsky, is a search for "the" fundamental "algorithm" behind language. The video mentioned above gives a taste of the formalism invented by Chomsky to "parse" speech in a way that supposedly doesn't depend on the language being used.

The "innate" capacity is distinguished from 

1. The perceptual systems that deliver stimulus to the brain - sounds or sights - that are the raw material to for language input and 
2. The physical mechanism of speech
3. Any particular language. Chomsky maintains that a baby starts to distinguish the unique features of his parent's language very early - possibly in the womb, but that any baby can learn any particular language.

Chomsky has had over 1/2 a century to flesh out these ideas, to defend them and turn them into something useful - a basis for a "Scientific" theory. The above presentation leads me to believe that he hit a dead end.years ago.

One positive aspect of Chomsky's approach is that he freely admits that there is a lot more going on in the mind besides language. One misses this kind of humility in S&E.

EVOLUTION

Whenever anybody invokes evolution to support or explain any aspect of his theory, red flags should go up. Chomsky is no exception. He prefaces his ideas on evolution by dismissing the public's "pop" idea of how evolution works, but it's hard to see how his own ideas differ. He still seems to be wedded to the idea that any aspect of an organism (phenotype) must be explained in terms of an adaptation beneficial for survival. This backs him into a corner where he feels that he must account for his (hypothetical) unique language capability in terms of what amounts to "pop" evolutionary theory. By his own (apparently unnecessary assumption), this capability must have popped into existence suddenly, "out of the blue" and not through the assembly of gradually-acquired mutations (such as increased area of the cerebral cortex).

One of Chomsky's key claims is the insistence that language is unique to humans, without parallel or precedent in the animal kingdom. Another related (but strangely irrelevant) claim is that language is the necessary and sufficient evidence of abstract thought. Thus, for example, he assumes that archaeological evidence of symbol making are evidence of language. This forces him into making silly statements such as his claim that dogs can't have ideas or make generalizations. This is only true if you define ideas and generalizations as language, which is preposterous.

He claims (needlessly) that the innate capability of language appeared recently and suddenly in evolutionary history, possibly by a single mutation. Obviously, language would be not much use to the single person who had this mutation. Of course, a mutation will quickly spread through a population, even if its "survival advantage" is neutral. If you think carefully about this, the argument that this capability appeared "suddenly" becomes meaningless. Any mutation can spread through a population after appearing only once (red hair for example). "Selection" will operate against the mutation only if it's harmful. Chomsky introduces evolution to "sound scientific". It's irrelevant to his theory.

The frog's eye generalizes what is perceived and sends a signal that amounts to "food" or "danger" to the frog's brain. That's analogy making. S&E is founded on analogy and has no need for anything to pop into existence suddenly with humans. Human analogy making can be "explained" by gradual improvement in the ability to make good analogies (the "Essence" in "Surfaces and Essences), which could very plausibly be associated with growth of the amount of grey matter devoted to analogy making throughout the brain.  Chomsky has no idea what the language "capability" is working with (certainly not analogy), so his theory is built on a huge, sudden, mysterious mutation. This isn't much different from a "Gift from God".

ACQUISITION OF LANGUAGE

Chomsky has been roundly criticized by his failure to account for the way real humans acquire language, yet he admits that any linguistic theory must fully account for how linguistic ability develops in humans. Listening to Chomsky, one gets the idea that it's been a long time since he's had a conversation with a two-year-old. S&E claims that children learn language by analogy - copying speech patterns and learning the situations where these speech patterns are appropriate, then (automatically and somewhat mysteriously) developing vast stores of analogy that allow them to use the same or similar noises to apply to "similar" situations. This makes a lot of sense and agrees with what we see in actual children. Chomsky's theories about how children learn syntax, especially how they form sentences into words, seems like a wishful fable in comparison.

REDUCTIONISM

Reductionism is still at the basis of so much "scientific" thinking that it's hard to spot in Chomsky's thinking. However, it's clear that his project is to reduce language to something simple and almost "atomic". In fact, he uses the example of chemistry, which was initially a project to simplify the principles and identify the underlying "rules". The idea that all phenomena can be "reduced" to effects that emerge from the actions of simpler, more fundamental systems has been discredited in the 21st century. We no longer see the need to "boil down" brain capabilities to "computations" in a meat computer between our ears. They can be studied on their own. There is nothing stopping us for explaining or describing these capabilities in terms of more fundamental processes in the brain, but we need not assume that such explanation is possible "in principle".

Chomsky is clearly in awe of the "hard sciences" and hopes that linguistics will evolve in that direction. However, he admits that that is not actually happening. In fact, the attempt is falling apart, partly due to its premature efforts to find formal "laws" that simplify the staggering quantity of data it tries to "explain". Linguistics is more complicated than the phenomenon it tries to explain/

One suspects that perceptions and analogies are the very "atoms" Chomsky is looking for to form a foundation for a theory of linguistics. But it's too late. He's explicitly rejected this option and clings to his theory that, by his own admission, is built on mysterious interfaces between language and the real world.

MATH ENVY

Chomsky's attempt at formalizing the "algorithm" at the heart of language capability fails at the very outset when he confuses the idea of a "set" with "category". As S&E so clearly demonstrates, the categories appearing in language are fuzzy, overlapping and sometimes surprisingly arbitrary. But  in the presentation (above) Chomsky explicitly says that any object cannot belong to two categories. Chomsky's categories can't overlap. Chomsky explicitly relates his ideas to those of Plato, which assume "pure" categories, like a "real" circle in the mind when an imperfect circle is perceived in the real world. It is not surprising to find Chomsky repeatedly stumbling into referring to "real" categories out there in the world. A "thing" belongs to a "real" category - language simply manipulates "real" categories. Language doesn't consist of forming categories on the fly by means of analogy. All this is clearly wrong and results from the hidden analogy between the brain and a machine. Computers actually do rely on crisp, clear categories like those imagined by Chomsky.

APPLICATIONS

To my mind, the biggest difference between S&E and Chomsky's theories is in the area of application. One cannot read S&E without gaining a whole new insight on language and the workings of the human mind. Applications abound, from teaching children in Kindergarten to analyzing the bad ideas that cause wars. There is no hint of application if Chomsky's linguistics in the video mentioned above, nor have I ever heard mention of a practical application. Failure to apply in the case of children has been noted.

I'm also reminded of a conversation I had with a young lady studying linguistics in University. She was totally unable to explain what her subject was about. This echos the situation with students studying philosophy or string theory. These are all subjects "about" themselves.

CONCLUSION

There is some value in going over Chomsky's breakdown of the problems involved to see how S&E solves them. For example, S&E provides a seamless transition between perception and language. It provides a similarly seamless transition between non-human "thinking" and human thinking. Chomsky is mystified by the problem of "free will", which boils down to extending our idea of thinking to cover action. This seems to pose no problem in S&E but is a huge problem for Chomsky, who regards the idea that thought and action are the same kind of thing as a throwback to behaviorism.  In fact, Chomsky himself has a very hard problem with categories and analogies even though he can't seem to talk for five minutes without supporting his own projects by means of analogy.

S&E would seem to sweep Chomsky linguistics aside. Whether or not S&E is "true" in detail, it is clearly a more useful and productive way to think about language. Chomsky's linguistics is already under attack for reasons that have nothing to do with S&E (it simply doesn't work), but S&E deals many of the mysteries and unsolved problems of S&E. To my mind, there is not much left of interest.

It should be remembered that both S&E and linguistics leave aside the realm of human thought that is not connected to language. Generally speaking, all the capacities associated with the "self" are presumably inbuilt capacities of the brain which remain intact even if speech capabilities are lost (for example, by a brain lesion in the left brain). While study of language teaches a lot about the human mind, it's not the whole story.

Letter to John Heerema, Sept 24, 2016

I am particularly attracted to the idea of the "fractal", which is a scale-independent repetition of a theme. Nature seems to "discover" a few patterns then use them in endless patterns and variations. This is one of the building blocks of what is taking shape as a "theory of mind". To build a "super mind", it would make sense to design the largest scale of this mind in the cloud to tap the computational power of millions of volunteer machines (as in the cloud project to classify galaxies).

However, perhaps this as already been done. At least to serve in my "thought experiment", Google may be enough. The "theme" that has me fascinated is the mapping of one brain state into another, using something similar to the Shrodinger Wave Equation. That equation describes the state of the universe in the "next instant" as a function of the state of the universe in the "previous instant". The mental equivalent would be the set of memes arising from the set of memes in the previous instant. At the level of the stream of consciousness of a single person, this might be a good way describing "thinking". There are good reasons to suspect that a single human brain can sustain several of these steams at once, perhaps millions. Daniel Dennett likes to use the metaphor of thousands of mini-minds "voting" on what should be brought to attention at any one moment. I think he's perhaps more right than he thinks. It's not just a metaphor. of On a smaller scale, this is a model of perception (what things "look like" or "sound like") and also a model of overt action ("why" we do some specific thing). 

Beyond the level of the individual, memes continue to map. In physics, we speak of a "core" theory, which is the Shrodinger Equation restricted to a specific domain, within which its results are "for all intents and purposes" accurate (for example, in some situations, you can ignore gravity). It's not hard to imagine the meme mapping that's going on in one person's stream of consciousness, but there seems to be no reason to restrict the idea to just one person or the lifetime of one brain. Entire societies "think" in this way, producing a new set of memes from the preceding ones every second. 

I'm fascinated by history. When I learn about World War I or the career of Napoleon, it seems to me that the entire world was "thinking" - following one set of memes to another in a crazy way that makes sense on some level. You can' speak intelligibly about these things using physics. They are more like thought processes and that's the way they are usually described.

What does Google do? You input a meme, in the form of a word, a question or a collection of words that, according to your best efforts, describe an idea. Google pops up with a few million "related" ideas ranked in order of "best guess". How is that different from thinking?

All this can seem interesting but vague. Dawkin's original idea of the meme as the mental equivalent of a gene didn't catch on because there was no way of defining what "meme" means. What is an idea? What is a perception? Like fundamentally new ideas, the word "meme" needed to be defined by using it. Hofstadter is the master when it comes to showing us how to use it.

Hofstadter is quite familiar with the word "meme" but talks about it using the more familiar term "category". His work can be seen as an effort to describe what we mean by a "category", and it turns out to have all the recursive, fractal, self-referencing properties of thought itself. In fact, Hofstadter seems to claim that thought is "nothing more" than forming categories.

Hofstader's work lends the requisite elegance to "meme" theory. He thinks (and he may be right) that meme mapping completely accounts for consciousness. Once you "grok" Hofstader's way of thinking about thinking, lots of the literature on the mind seems to suddenly make sense. I'm currently reading Gilbert Ryle's classic "The Concept of Mind" (1947) which is one of the best-known assaults on the Cartesian theory of mind. Ryle sits right at the edge of a revolution. He's trying to teach us how to speak sensibly about thought processes as being in the same category as "overt action". Seen through Hofstader lenses, his argument convinces me that actions are memes. AHA! 

To my mind, these insight is at least as productive as Newton's discovery of calculus. Applying meme theory to its own emergence, we see that it's not the product of a single mind. It's popping into existence all over the place. Like calculus and relativity, it was there all along, waiting to be discovered. I'm just running with it to see where it leads

P.S.

After sending this note off to John, I let my dogs in from outside. I watched the two of them "calculate" how to get in the door while dealing with the distraction posed by he competition between them. It struck me that dogs (and animals in general, even fleas) map perception into action more or less directly ("instinctively"). That is, meme theory applies to everything with a brain, no matter how simple that brain may be. This is thoroughly consistent with the fractal nature of mind.

Idea #2 - how does Dawkin's original idea - survival of memes - work in to all of this? Can we make sense of the "survival" idea with respect to memes?