The Language Bottleneck
Published: 2019-08-21 . Back to ≈

“I am God, and I have no need to think. Up to now I’ve never thought, and I’ve never felt the need, not in the slightest. The reason human beings are in such a bad way is because they think; thought is by definition sketchy and imperfect—and misleading. To any thought one can oppose another, obverse thought, and to that yet another, and so forth and so on; and this inane cerebral yakety-yak is about as far from divine as you can get.”

So we meet the protagonist of Giacomo Sartori’s novel, “I am God.” An unconventional God, to be sure.

God has no need to think. Perhaps it isn’t what Sartori had in mind, but I’m taking this to mean that God has no need for language. And more to the point, he finds language to be unfit for a god. The use of language entails compromise and self-contradiction. (Why do we find him using words here? He is a keeping a diary–just one piece of his descent into the human sphere brought about by an ungodly obsession with an ungodly biologist. Read the book if you want to know more.)

I’m starting with this quote because it communicates that basic idea that language involves limitation. A God of infinite mind would ostensibly have no hope of expressing his mind with a finite stream of words. But perhaps a better word than limitation is compression. Yes, language is limiting. But humans have limited minds occupied by ideas and concepts of finite size. Often it seems perfectly feasible for us to take those ideas and represent them with language. That’s what I’m trying to do now, of course.

This notion perhaps only becomes interesting when one asks whether a bottleneck (mind –> language –> mind) is a necessary precondition for communication. Or thought. A God would surely think. But use language? Humans communicate via creating vibrations with our vocal chords in order to transmit soundwaves through the air, to be received by the ears of others. But could life have developed under other conditions to have communication over higher-bandwidth channels? Might it still?

On the other hand, unburdened from the limitations of language, does thinking perhaps run into difficulties? Is this limitation responsible for the successes of human civilization?

The Language Bottleneck

As I sit here writing this blog post, I’m listening to music. Just writing this much allows my reader (real or imagined) to reproduce my experience with a certain level of fidelity—in this case, likely very little fidelity at all. In reducing my experience to effectively a single word, “music,” I have stripped away all but the broadest strokes of definition. If I add more detail, e.g., “I’m listening to Jazz music” or “I’m listening to Bossa Nova,” I allow the reader to gradually reproduce the experience with finer strokes. Greater skill or meticulousness would naturally allow me to convey various aspects of the experience with yet finer detail.

Now, let’s imagine that I could share my experience with the reader directly through some kind of mind-meld or telepathy. Suddenly, there would no longer be a need to reduce the experience down to words. All such efforts would convey only a crude message when compared to the direct sharing of experience.

Absent of such abilities, our actual conduit of communication—sounds made with the mouth—is of a limited bandwidth, requiring our thoughts to be compressed down to a form suitable for that conduit: words and sentences. Through biological learning and cultural evolution, humanity has learned to utilize that conduit with remarkable efficiency. The structure of our language reflects the statistical regularities of our experience in a way that allows us to convey thoughts and experiences with remarkable fidelity, often using only a small number of words. If I were to say, “I pet the ginger tabby cat on the side of the trail,” you would probably be able to imagine the experience in a way that would be very much alike in quality to my own memory of the event.

The Mind Bottleneck

So far, I’m not sure that I’ve said anything particularly controversial (see the next section for whatever controversy there might be). What I intend to be the interesting observation is this: that without some kind of communication bottleneck—in our case, stemming from the fact that humans happen to have mouths and not high-bandwidth antennae—humanity might never have developed literature, poetry, or even science.

Take the last item, science. Science is a kind of language. Specifically, scientific law is the language of a subculture obsessed with saying only that with is falsifiable, yet not falsified. Where everyday language reflects cultural concerns of everyday approximations, scientific language reflects the concern of the scientific endeavor with precision, objectivity, and empiricism. The difference between mathematical expressions of scientific theory and, say, English, is a cultural difference beyond all else. The difference in the symbols and syntax used is superficial. Put differently, all of science is just a model that we have created which can be used to make predictions about the aspects of our experience which we call objective. (By objective, we can’t mean anything more than the aspects of our experience that transcend our individual subjectivities. Things that you and I and everyone else can agree on. That doesn’t make those things not subjective, it just makes them a subset of the subjective.) The scientific models, themselves, also live in our minds. Scientific theories, articulated mathematically or otherwise, are a reduction of those models into a code, which is encoded by one mind and decoded by another.

Putting the matter in this way has interesting ramifications for topics like general artificial intelligence. Philosophers like Nick Bostrom seem to be fond of placing human intelligence on the lower end of a spectrum that stretches out to the right into unknown horizons. Indeed, it isn’t difficult to imagine lifting certain constraints on the human intellect. Example: imagine a human with an ability to simultaneously think and reason about 1 billion “chunks” compared to a human’s paltry working memory of half-a-dozen items.

But what about the cultural aspect? If an intelligent super-being could simply learn or intuit the behavior of complex systems like the climate or the solar system, would it need to simplify and reduce those behaviors into a language like our science? If so, what would that language look like to us? If not, how would the lack of a scientific language in turn limit that intelligence?

An Ontological Detour

My basic idea is that language is the byproduct of having a limited channel—our mouths and ears—by which to share thoughts. This limitation has resulted in a neural pattern alike in function and perhaps form to a bottleneck, by which our experiences/thoughts can be translated into words.

To put this idea into relief, let me contrast it with another idea—an unexamined assumption for some and a serious philosophical proposition for others—that words refer directly to the essence of a thing, itself, an essence which might best be referred to as the thing’s Platonic Form.

The Platonist assumes that an ordering principle exists objectively in the world. The word ‘frog’ is meaningful because there exists an essential form or idea of the frog which is accessible to the mind apart from a direct encounter with a specific instance of a frog. Moreover, the sentence, “The frog is in the well,” has a sort of objective meaning that inhabits the very words themselves. More so, the sentence, “My pet frog Ivan is in his terrarium at my house.” The nouns of this sentence are taken to have very specific and real referents, and the verbs to declare a state about those objects, such that the sentence as a whole has a solid and robust meaning that is beyond subjectivity.

In contrast, the idea of language as a low-dimensional representation of experience dispenses altogether with the ontologically questionable idea of the Form. The word frog is meaningful because of the neural apparatus of the brain, which repeatably and reliably reduces a certain regularity of visual stimuli to a single label: frog. (Of course, the original visual regularity stems from the regularity of the universe itself.) Without the neural circuitry for encoding into and decoding out of language, the words “The frog is in the well” are merely sounds—just as the internal representations of a bottleneck layer of an artificial neural network are merely meaningless numbers when separated from the ensuing decoding or classification layers. The words must be taken together with the neural system as a composite whole. Moreover, the image of the frog that presents itself to me, with elements of color and divisions of form and function that I can immediately perceive, exists nowhere in the universe outside of my mind or some mind (or perhaps information processing system).

The idea of the Form is perhaps most compelling when applied to abstract ideas or entities in mathematics. For example, one might argue that statements like “something exists” or 1 != 0 clearly correspond to fundamental truths that transcend experience; that it would be wrong to say that they merely characterize or communicate experience.

I would counter as follows: What makes these statements seem “axiomatic” is that it is impossible to imagine (an experience) the contrary of what we mean when we say either of them. It’s impossible to imagine being dead, such that no experience exists. It is impossible to imagine one thing being the same as zero things. But this second impossibility depends on ability to count, which is a way that our minds organize the world—the rudimentary beginnings of a prediction engine which nonetheless had to develop over time. The idea of zero was a late arrival to the scene.

In fact, we have to have a lot of language to even say something like 1 != 0, and have it mean anything. Like the idea of equality or identification. This isn’t just a trivial idea in itself. Our language has evolved many different concepts of equivalence, each suited to the particular purpose or application.

So, is 1 != 0 a fundamental fact of reality? Or is it the consequence of a system of language which has slowly evolved to help us describe, predict, and reason about our experiences? You decide!

Why Write?

When seen this light of the language bottleneck, I think that words can take on a tantalizing kind of dynamism (which the confines of Platonism would never allow). For example, we can readily admit that the same phrase has different meanings to different people, without feeling that we are doing violence to the notion of meaning. But more than this, we can fully appreciate the way in which the masterful application language can set the listener’s mind into multifarious modes of resonance or reactivity. The skilled drummer uses his understanding of the tension and contour of the stretched surface of the drum to direct his blow so as to produce the correct tone and timbre of sound. So the skilled user of language uses understanding of connections in the mind of the reader or listener to construct a sequence of blows, perhaps with carefully measured timing and intonation, that will set the whole thing ringing with powerful images and sensations.