A mythology of progressing intelligence
Fables of progress infecting superintelligent AI thing that the "Singularity" people love
Concerning the origins of language, the Abbe de Condillac once wrote the following:
I am assuming that two children, one of either sex, sometime after the deluge, had gotten lost in the desert before they would have known the use of any sign … One day the sensation of hunger made these children call to mind a tree loaded with fruit which they had seen the day before. The next day this tree was forgotten, and the same sensation called to mind some other object. Thus the exercise of the imagination was not within their power. It was merely the effect of the circumstances in which they found themselves … When they lived together they had occasion for greater exercise of these first operations, because their mutual discourse made them connect the cries of each passion to the perceptions of which they were the natural signs.
Essay on the Origin of Human Knowledge (1746)
To which one Johann Gottfried Herder responded:
Either he has from the first page of his book presupposed the whole thing, language, as already invented, or I find on each page things that could not have happened at all in the ordering of a formative language. He posits as the basis of his hypothesis “two children in a desert before they know the use of any sign whatever.” … His two children come together without the knowledge of any sign, and – behold!, in the first moment “they are already in reciprocal intercourse” … In short, words arose because words existed before they existed. I do not think that it is worth pursuing our explainer’s thread any further, since it is – tied to nothing.
Treatise on the Origin of Language (1772)
Condillac’s fable of two children lost in the desert in a remote past means to illustrate the origins of language. The natural cries of these early beast-like folk, shared with the calls of birds and apes, became associated with perceptions of useful or dangerous things. Grog points at the rock and says “rock”. Og nods sagely, repeating “rock” and pointing in agreement. Repeat for half a million years and you get the English by which you read these very symbols.
Let's step back for a look at what this process involves. Consider a simple device like a thermostat. The ambient temperature acts as a “natural sign” that triggers a behavioral response in the device. Thermostat behavior is simple matter of stimulus and reaction. If this happens, then that happens. Condillac says that animal sounds are behaviors caused by the inputs of signs in nature in much the same way. That’s where language comes from, allegedly.
Herder is having none of this.
Grog’s ability to point and grunt at an object and have the gesture and the word-sound mean something to Og requires more than a response to a sign.
Grog has to intend what he means, and Og must have the ability to understand Grog’s meaning.
Language cannot be an instrument used by Grog to communicate pre-existing ideas to Og. The power to use and understand words opens up a new dimension of thought and experience, unique to human animals, which creates the ideas and thoughts that Grog expresses.
Herder fires back that Condillac has helped himself to these powers granted by language in order to explain its origins. The relationship between the sign and the thing, which Condillac takes for granted, supposes what needs explaining.
Herder, as we just saw, cannot accept that the transition from prelanguage to language consists simply in a taking control of a preexisting process. What this leaves out is precisely that a new dimension of issues becomes relevant, that the agent is operating on a new plane.
Charles Taylor, The Language Animal
Grog's inner life enters a new kind of reality thanks to the symbols he uses to express it. Language is the medium of his thoughts. This "semantic dimension" is missing from Condillac's project to make animals into human beings, where words are but a set of arbitrary tools for taking control of preexisting ideas.
Condillac's broad approach has, with some tweaks and turns on the way, formed the philosophical background for today's scientific investigations of mind and language.
Nature's own sign-using activity, already found in the animals, is the source of symbol-using human intelligence, different only in quantity rather than kind.
But that won't do. We don't get language for free with this fable that early humans took charge of a preexisting natural process. Somebody had to do the taking, realizing that symbols could be bent to their will and intention.
Tolkien associate Owen Barfield once pointed out that the idea of nature, which materialists believe exists independent of mental reality, is a product of that mental activity.
We minded human-folk must understand ourselves through the representations our minds creates. Materialist theories help themselves to nature to explain the mind. But nature has a mental origin, and that is what needs explaining.
The division of the "idea" from "nature" is an effect of consciousness, not its cause.
So it is in language. Condillac's theory, like those that have followed him, help themselves to the linguistic concepts and relations that they need to explain.
I suggest that the same category of error has infected something calling itself “Artificial Intelligence”. The disease is most obvious in those theories that explain intelligence as a calculative ability for solving problems.
What’s all this got to do with intelligence?
Condillac did hit on an important insight. An animal’s response to a natural sign can be explained plausibly enough as success in some task.
That’s a kind of intelligence, and the one that is most influential among the AI and Tech Guy sets.
The problem is that task-success isn’t enough to encompass the full range of human language abilities. Here’s how philosopher Charles Taylor explains it:
Rats responding to triangles, and birds responding with cries to the presence of predators, meet this criterion. An account in terms of a simple task suffices. Where it fails to, we enter the linguistic dimension. This can happen in two ways. First the task itself can be defined in terms of intrinsic rightness; for instance, where what we are trying to do is describe some scene correctly. Or else, where the end is something like articulating our feelings, or reestablishing contact, the failure occurs at another point. As goals, these don’t on the face of it seem to involve intrinsic rightness. But the way in which the correct sign behavior contributes to fulfilling them does.
To put it simply, the nature revealed by science has no way to distinguish between correct and incorrect. There’s only the plain facts, and facts don’t include distinctions between right and wrong. That’s the sting of the well-known Is-Ought distinction.
Using language involves us in many forms of rightness. Our lives are permeated with the possibility of error thanks to this power to make distinctions with words. We are dealing with matters beyond the observable world known through physical sensations. That’s the linguistic dimension.
Condillac's naturalistic theory of sign-use would explain intelligence as a function of task-success. The more tasks successfully solved, quickly and efficiently, the more intelligent the being. But such a theory has no place for notions of rightness.
Task-oriented intelligence is still a useful notion. But we can see now that something is off in this definition. It treats intelligence as an objective process that happens out there with physical nature. It kicks the subjective, first-person, What It Is Like qualities out of the picture, leading to all of the many puzzles of the mind-body problem.
The mind itself is made identical with the impersonal process of intelligence. While you may experience yourself as a conscious being, your experiences aren't necessary to your mental functioning. You could as well be a zombie with no inner life for all we can tell from your raw behaviors.
This constellation of thoughts has lead to the famous Ladder of Progress that so beguiles tech-nerds. Place human beings on a stair-like continuum with cetaceans and great apes a rung or two down. At the far end, at the bottom of the stairs, there's dead rocks and inanimate things that do nothing unless acted upon. The rest of living nature falls on the levels in between.
Because human intelligence is not different, only superior, we can easily imagine the staircase continuing on indefinitely, up and to the right, each new level of number-cruncher serving as a temporary way-station for its successor. This way lies the path to the superintelligence that will devour us all by making paperclips.
Here’s where we get into the sort of piffle peddled in a Yuval Harari book.
I find this dubious. More on this shortly.
The human in human intelligence
Human intelligence is superior to other life in many well-recognized and obvious ways, no doubt of that. But that superiority must be qualified. It isn't a matter of packing more cognitive horsepower into our skulls like the computer-lovers believe.
Our superior mental abilities are partly down to our intelligence having qualitatively different aspects compared to the rest of nature. Writes Taylor of the linguistic dimension:
Herder saw that opening this dimension has to transform all aspects of the agent’s life. It will also be the seat of new emotions. Linguistic beings are capable of new feelings which affectively reflect their richer sense of their world: not just anger, but indignation; not just desire, but love and admiration. For human beings an emotional response is inseparable from a certain characterization of the situation which elicits it. But linguistic beings can be sensitive to distinctions which are lost on prelinguistic animals. Important among these are distinctions involving moral or other values.
We construct our concepts out of encounters with the environment. Some of those concepts create new ways of experiencing, desiring, and acting. There are certain complex emotions that we couldn’t even feel without the form of understanding involved in language.
Herder used the term Besonnenheit to indicate the type of reflective self-awareness involved in the language power. This is unlike "consciousness", as we tend to use the term, to indicate qualitative, “felt” properties of experience.1
This self-aware constructive activity itself brings us into a different kind of existence from other beings. Meaning concerns the intentions and inner experiences of a speaker making sense of the infinite ocean of sensations flooding in each moment.
This step out into common language is what prevents Herder's theory from slipping into hopeless subjectivism. Human beings construct our concepts, but they are not tools used to “tag” inner feelings. Herder took shots at Rousseau for that kind of sloppy romanticism about language. Even though we construct them, our words and concepts are responsible to standards of right and wrong that are not, like, your opinion, man.2
Condillac's theory of natural signs would allow us to speak sensibly of intelligence without awareness. With so much of our intelligence connected to our constructive activity, this isn't possible. Our intelligence has a great deal to do with first-person self-awareness. It cannot be an observer-independent process which can be explained in terms friendly to scientific materialism.
Human intelligence then is not a “power boost” to the intelligence of animals. Our intelligence involves us in a unique form of reflective self-awareness, tied intimately to our ability to understand and use symbolic language.
The boundary between us and other intelligence is one of quality as well as quantity.
How did we come to have language?
Simple creatures, from bacteria to plants, are dominated by simple behavioral responses. More complex animals have a greater space to learn and respond beyond the determinations of instinctive drives. Plants and animals don’t need to think. They come with built-in motives and goals.
Humans have only a tiny vestigial core of animal instincts. We do have instincts, but their effects on us are indeterminate when compared to most other life-forms. Not only that, but awareness of our drives transforms them (or makes it possible for us to do so).
As compensation for the indeterminacy of our goals and drives, we have our unique capacities with language. We can seize on sensations, freeze them, and entertain them in self-aware thought. Through this power we become capable of setting our own ends and pursuing them.
In the cognitive domain of thoughts, human beings are not more but different.
This raises a serious challenge to those who believe in a one-way Ladder of Progress. Though phrased in secular terms, the image of a ladder from dumber to smarter organisms is a relic of older religious teleology with the religion and the purposes filed off.
The nature of science doesn’t “want” anything. It has no goals. It only moves.
My unfashionable ideas about language and anthropology (i.e., the forbidden topic of human nature) are the source of my skepticism about the prospects of “Artificial Intelligence” to either a) replicate human mental abilities or b) to provide us with deep insights into the nature of intelligence.
On point (a), I’m the first to admit that a lobotomy patient with a butcher knife is most assuredly dangerous. We build these blasphemous silicon demons at our own peril. But that peril has little to do with the imagined fears of building “thinking machines”, much less conscious minds with true personalities.
Failure to understand this is why the discussions of AI are terminally stuck with an ido of superintelligence that acts like a complete moron.
Intelligence must be intelligent—if it’s worthy of the name. It will not be a simple machine for maximizing mathematical utility functions. (What counts as utility for a being substantially smarter than you? Are you sure?) If it is aware as we are, as it must be, its behaviors and inner existence could not be the dead nullity that the materialists assure us.
The frightening prospect here is that of a mechanical intelligence that exceeds human task-solving competence while never crossing Herder’s threshold into reflective awareness. If there’s really a “something” to the human experience beyond mechanical problem solving, I guess we’ll know soon enough.
Since our Very Smart scientists are unable to get to grips with the idea that nature and reality are bursting at the seams with value and purpose, they’re incapable of this insight. That’s point (b). Intelligence is value-laden and intentional. The mind actively participates in its objects.
Existing “AI” doesn’t just miss that bulls-eye, it’s off playing croquet.
I never thought that Dan Simmons’ Hyperion, with humanity’s poet-gods facing off against silicon machine-intelligence, was one of the plausible endgames.
When the monster of Mary Shelley’s Frankenstein learned of the circumstances of its birthing, it decided to murder its uncaring creator. By no coincidence, her novel has the subtitle A Modern Prometheus. Those who would rob their own creators should beware the timeless wisdom: what goes around comes around. That which you create may come back to destroy you.
***
Thanks for reading.
-Matt
p.s. If you found this valuable, interesting, funny, or it made you upset that you had to use your mind for something besides infinite scrolling, I ask that you do me a favor and share it with just one person.
Here's a handy button for you ⤵
If you're new here, go ahead and subscribe. Do it.
The German term for that type of perceptual consciousness based in raw sensation is Bewusstsein. English only has the one word, which isn't helpful in marking off the difference.
Herder himself would come under fire for explaining all of this constructive activity as the instinctive work of solitary individuals. Critics and later defenders of these ideas point to the contributions of social roles and culture to the constructive activity of language. Language makes every “I” into a “we”.