The Limits Of Language & Logic – Part 1
Words can be spells. Language is perhaps the most wondrous and dangerous of our discoveries and technologies, a magic that must be respected and used very carefully.

Joefrei, CC BY-SA.
As kingfishers catch fire, dragonflies draw flame; As tumbled over rim in roundy wells Stones ring; like each tucked string tells, each hung bell’s Bow swung finds tongue to fling out broad its name; Each mortal thing does one thing and the same: Deals out that being indoors each one dwells; Selves — goes itself; myself it speaks and spells, Crying Whát I dó is me: for that I came.1 (Gerard Manley Hopkins)
Like a modern Hephaestus forging life in rhythmic, ringing blows, Hopkins beats life into each word and line. Poetry as incantation. Words can be spells.
Language is perhaps the most wondrous and dangerous of our discoveries, a magic that must be respected and used with care. Our creation myths know this: Ptah’s divine utterance in the Memphite Theology; “God said, Let there be light”; the Greek Logos, in which speech is generative and gives the cosmos its order. The breath of language is the father of reason.
Words are tools. Language can give us the illusion that it represents reality; the better the performance, the closer the approximation. But we should not mistake the game for the deeper, unspeakable reality. As Bruce Lee put it: “It’s like a finger pointing away to the moon. Don’t concentrate on the finger or you will miss all that heavenly glory.”
“Words bend our thinking to infinite paths of self-delusion, and the fact that we spend most of our mental lives in brain mansions built of words means that we lack the objectivity necessary to see the terrible distortion of reality which language brings.” (Dan Simmons, Hyperion)
``
Introduction #
Our use of language must begin with an understanding of its risks — delusion, confusion, and the belief that abstract concepts are real things — and of its irreducible limits. Language is a game of context, and a wonderful one when played well, but our problems with it can descend into a collective mental illness.
Logic has limits of its own. Gödel showed that truth cannot be fully determined within any sufficiently powerful self-referential system. Recursive logic is a useful language; that does not make it truthful.
In this Part we consider the philosophy of language: Wittgenstein’s journey from pictures to games, Turing’s test and Searle’s room, and Chomsky’s universal grammar. In Part 2 we turn to the limits of logic itself, and to what lies beyond both.
“Without philosophy thoughts are, as it were, cloudy and indistinct: its task is to make them clear and to give them sharp boundaries.”2(Wittgenstein)
Like DNA, languages carry the remnants of our evolution. Words are extraordinary compressions of history and infectious carriers of culture. Every time I say “goodbye” I wish that God be with you, even though I am not religious and did not intend it. Words can be lyrical, assonant, dissonant; they may even sound like what they name.
“Philosophy is a battle against the bewitchment of our intelligence by means of language … The philosopher’s treatment of a question is like the treatment of an illness.”3(Wittgenstein)
For such little things, words do a tremendous amount of work. These superhero sherpas carry whole worlds on their backs.
Could mortal lip divine The undeveloped Freight Of a delivered syllable ‘Twould crumble with the weight. (Emily Dickinson)
Each word is a sign and a seed, ready to grow into a many-branched tree. Their strength is in how much we can pack into them and how gregarious they are, flocking together in great murmurations. Words working well together are a fine example of algorithmic efficiency.
Wittgenstein’s Language Games #
Wittgenstein began in the analytic school and changed his mind about language more radically than almost any other philosopher.
The early view(Tractatus Logico-Philosophicus, 1922).
Language is a picture of reality. A proposition has sense if it pictures a possible state of affairs, and it is true if that state of affairs obtains. Whatever cannot be pictured — ethics, aesthetics, the mystical — cannot be said, though it may showitself.
“Whereof one cannot speak, thereof one must be silent.”
Consider: The giraffe is lying on the couch. On the early view this has sense because it pictures a state of affairs we can check (though note that it might also mean the giraffe speaks falsehoods from the couch). Now consider: God is love. This pictures nothing we can point to. On the early view it has no sense — though Wittgenstein would add that what it tries to say may nonetheless be of the greatest importance.
“Language disguises thought. So much so, that from the outward form of the clothing it is impossible to infer the form of the thought beneath it, because the outward form of the clothing is not designed to reveal the form of the body, but for entirely different purposes. The tacit conventions on which the understanding of everyday language depends are enormously complicated.” (Wittgenstein, Tractatus 4.002)
What’s in a name? That which we call a rose By any other name would smell as sweet. (Shakespeare)
The later view (Philosophical Investigations, 1953). Language is a tool for playing games. There are many kinds of language game, each with its own rules; the meaning of a word is its use within the game being played. Words have no inherent meaning; context gives it to them — a thought that echoes the Buddhist doctrine of śūnyatā (emptiness) and the ecological concept of affordance.
Now The giraffe is lying on the couch carries no independent truth. “Giraffe” need not refer to an animal at all; depending on the game it may be shorthand for something abstract, comic or artistic. And God is love is perfectly intelligible within many games — as a claim about a benevolent power, as parody, or as the coded shorthand of a community living under a regime that forces the phrase on its people while practising hate in that god’s name.
| Early Wittgenstein | Later Wittgenstein | |
|---|---|---|
| What language is | A picture of reality | A tool for social activities |
| Where meaning comes from | Correspondence to facts | Use within a language game |
| Essence | Logical atomism: complex propositions reduce to simple ones | Anti-essentialism: no fixed essence; meaning is fluid and contextual |
The later view allows a word to hold two opposing meanings at once, depending on the game — a superposition of meaning. It also explains why language is so hard for computers: meaning is relative, contextual and always evolving.
Terrence Deacon’s The Symbolic Species develops the same insight: human language is a symbolic system in which words do not correlate directly with objects but stand for them through shared convention.4 Whether current AI has the cognitive architecture that produced human symbolic thought is an open question; evolutionary pressure on AI systems may yet produce deeper symbolic dexterity.
Turing’s Imitation Game #
“I PROPOSE to consider the question, ‘Can machines think?’ This should begin with definitions of the meaning of the terms ‘machine’ and ’think’.”5 (Alan Turing)
Turing immediately noted that defining the words by common usage would reduce the question to a Gallup poll, “but this is absurd”. He replaced the question with a game:
- An interrogator communicates by text with two hidden participants, one human and one machine.
- The interrogator tries to determine which is which.
- The machine tries to be taken for the human.
- The interrogator may ask anything.
Turing predicted that by 2000 a machine would leave an average interrogator with no more than a 70% chance of a correct identification after five minutes. He also wrote:
“The original question, ‘Can machines think!’ I believe to be too meaningless to deserve discussion. Nevertheless I believe that at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted.”
The test is objective because it does not rely on the interrogator’s view of personality or appearance. Turing dealt briskly with the theological objection (why should elephants, or any animate being, not also have souls?) and with what he called the “heads in the sand” objection:
“We like to believe that Man is in some subtle way superior to the rest of creation … I do not think that this argument is sufficiently substantial to require refutation. Consolation would be more appropriate.”
Turing also foresaw a machine that observes its own operations and learns from them: “If it is accepted that real brains, as found in animals, and in particular in men, are a sort of machine it will follow that our digital computer suitably programmed, will behave like a brain.”
His non-discriminatory stance against unjustified human exceptionalism remains the right starting point, both for assessing machine intelligence and for assessing the intelligence of other species.
``
Searle’s Chinese Room #
“Imagine a native English speaker who knows no Chinese locked in a room full of boxes of Chinese symbols (a database) together with a book of instructions for manipulating the symbols (the program). Imagine that people outside the room send in other Chinese symbols which, unknown to the person in the room, are questions in Chinese (the input). And imagine that by following the instructions in the program the man in the room is able to pass out Chinese symbols which are correct answers to the questions (the output).”6 (John Searle)
Searle’s Chinese Room Argument (CRA) is meant to show that proficiency at a language game is not understanding. The man passes a Turing test for Chinese without understanding a word; since a computer has nothing the man lacks, neither does the computer. Searle concludes that simulation is not understanding, that intentionality is a product of the causal powers of brains, and that algorithms are not sufficient for intentionality: “we cannot get from the syntactical to the semantic just by having the syntactical operations and nothing else.”7
Let us give Searle his due. An argument that has annoyed philosophers for more than four decades is doing something right, and the CRA is a useful prompt for asking what general intelligence would need. The standard replies — systems, robot, brain-simulator, other-minds, many-mansions — are well known and Searle defends against them ably.
The stronger objections are these.
It is not about Turing’s test. The Turing test does not test for understanding or intentionality, and Turing never claimed that passing it conferred either. He deliberately avoided “what is thinking?” because the question is indeterminate. Searle changes the test to one of understanding, leaves that undefined, then equates it with intentionality, which he also leaves amorphous.
It is circular. The argument runs: (1) the man in the room does not understand Chinese; (2) a program is like the man; (3) therefore no program can understand Chinese. Premise (1) is the conclusion in disguise. Searle presupposes that understanding requires something intrinsic that programs lack, then defines the man’s rule-following as not-understanding because it is rule-following. To escape the circle he would need an objective definition of understanding not biased against programs, and a test for it independent of the process that generates the answers. He offers neither. A program that passes the Turing test fails the Searle test — but how would anything pass the Searle test? We are not told.
It ignores learning. Understanding begins with mimicry. When Young and Champollion began deciphering hieroglyphs, they worked from syntactic skill in other languages and the parallel texts of the Rosetta Stone — an essentially algorithmic operation a computer could perform. Did they not understand Egyptian until some further non-algorithmic ingredient appeared? Every child learns a language by first getting a feel for grammar and imitating sounds. Imitate, assimilate, innovate (Clark Terry).
It treats understanding as binary. Understanding is a spectrum and operates on levels. We can catch the feeling a poet intended without knowing why; we can get the gist of an essay without knowing every word. Searle concedes this in passing, but the CRA needs understanding to be all-or-nothing.
Hofstadter is blunter still about Searle’s method:
“Searle gives the appearance of tossing off these humorous images [toilet paper, beer cans] lightheartedly and spontaneously, but in fact he is carefully and premeditatedly instilling in his readers a profound prejudice, or perhaps merely profiting from a preexistent prejudice.”8
Searle is right about one thing: the process of answering is not the reason for answering, just as a recipe is not the cake nor the appetite. But Turing never said an algorithm was both the process and the desire. To give Searle the benefit of the doubt, when we say “think” we usually also imply “understand” — words flock together — meaning something like having a framework in which to contextualise what is communicated and a motive for communicating. To insist that this must further imply a new and undefined concept of intentionality is a step too far.
“It is no secret. All power is one in source and end, I think. Years and distances, stars and candles, water and wind and wizardry, the craft in a man’s hand and the wisdom in a tree’s root: they all arise together. My name, and yours, and the true name of the sun, or a spring of water, or an unborn child, all are syllables of the great word that is very slowly spoken by the shining of the stars.”9 (Ursula Le Guin)
Semantic Shell Games #
“A problem that proponents of AI regularly face is this: When we know how a machine does something ‘intelligent,’ it ceases to be regarded as intelligent. If I beat the world’s chess champion, I’d be regarded as highly bright.”10 (Fred Reed)
Searle’s trick was to dispute Turing’s functional equivalence by swapping intelligence for understanding, and understanding for intentionality, without defining either. The danger is a semantic shell game: whenever one quality of understanding is shown to be present, the argument moves to a neighbouring concept, and so on indefinitely. This is the “AI effect”, and it is exactly what Turing’s test was designed to avoid.
``
The Buddhist doctrine of dependent origination applies to these concepts too. “Intelligence” and “understanding” are contextual, multivariate aggregates with no fixed essence. Trying to disprove one abstract concept by introducing a related abstract concept does not advance the argument; it creates a recursive loop and a defective language game.
If Searle’s intentionality requires identity and knowledge of other identities, then adults and older children have it, very young children perhaps not, and swarm insects may have it only collectively. Like an errant prophet, Searle leads us away from Turing’s question without delivering us to the promised land.
Understanding a language is more than linguistic competence; it is comprehension within a culture, emotional intelligence and the creative, authentic use of words. If large language models can solve new language problems beyond their training data and persuade an expert that they are fluent and sophisticated, it is hard to see on what principled basis we exclude them from understanding. Imagining an agent that simulates understanding without possessing it does not invalidate the possibility of understanding, human or digital. Without a test of comprehension, the CRA is specious and without predictive power.
People are not going To dream of baboons and periwinkles. Only, here and there, an old sailor, Drunk and asleep in his boots, Catches tigers In red weather.11 (Wallace Stevens)
Philosophers and linguists should pay more attention to poets, who convey meaning by bending and sometimes breaking the rules.
Chomsky’s Black Box #
Noam Chomsky distinguishes the unconscious internal language (I-language), which we cannot inspect directly, from E-language, its external expression. Linguistics, on his account, studies expressions, not language itself.
His central claim is Universal Grammar (UG): humans are born with an innate faculty for language, a set of principles shared by all languages — structure, word order, recursion — that lets children learn any language quickly from limited exposure. UG sets limits on how languages can vary and lets a child internalise rules before meeting them. It is genetic, inherited by all Homo sapiens, and on Chomsky’s view unique among animals: so-called sign language in great apes “is nothing like human language”, and complex language may never have existed even in Neanderthals.12
Over time, and perhaps because many exceptions to his universal rules were found, Chomsky narrowed UG to computational recursion — the capacity to combine finite words and rules into a theoretically infinite set of expressions, “discrete infinity”. Others dispute even that, arguing that language is a discrete combinatorial system rather than a recursive-embedding one.13
We have not found conclusive evidence of generative recursion in animal communication. But as Chomsky himself notes, all life is built from an older language of discrete infinities — the GATC of DNA, whose recursions have produced an endless diversity of forms. Human language is not the language of DNA, but it is a natural extension of the same deeper laws. It is another order higher than anything in the animal kingdom, and perhaps unique; it remains, first and last, an evolutionary product.
“There is grandeur in this view of life, with its several powers, having been originally breathed into a few forms or into one; and that, whilst this planet has gone cycling on according to the fixed law of gravity, from so simple a beginning endless forms most beautiful and most wonderful have been, and are being, evolved.”14 (Charles Darwin)
Chomsky’s writing is difficult, and his theory has shifted. His earlier generative grammar posited a surface structure (syntax, word order) and a deep structure (meaning) from which the surface is generated. His later Minimalism deprecates rules in favour of maximal efficiency with minimal structure. He has claimed that language arose through a single mutation yielding “unbounded Merge” around 60,000 years ago, conferring selective advantage in thought and planning before communication.
What persists across his theories is that language is cognitive (a product of mind, not merely environment), universal, generative and recursive, rule-governed below consciousness, minimalist, and an evolutionary discontinuity. He endorses Marr’s three levels of analysis — computational theory, representation and algorithm, hardware — and holds that AI’s reliance on ever-larger datasets neglects the implementation layer.15 On his view there is no algorithm for language, which echoes Searle.
To my mind it is simpler to let grammar and language keep their everyday meanings: language is a communication game between people, and claims about it should branch from frameworks that connect us to all life on Earth.
Birds do it, bees do it, Even educated fleas do it.16 (Cole Porter)
We know little of flea education, but bees waggle to tell their sisters where food lies, crows are exceptionally talented communicators and may even grasp recursion (Chomsky is unconvinced),17 and chimpanzees communicate about threats, food, status and greeting.18 AI may yet help decode something of dolphin speech.19 Linguistics has only recently begun to shake off its human-centrism, mirroring Wittgenstein’s own journey from language-as-truth-picture to language-as-communal-play.
Chomsky has his critics:
“But Chomsky is no Einstein. And linguistics is not physics. Unlike Einstein, for example, Chomsky has been forced to retract at one time or another just about every major proposal he has made up to his current research, which he calls Minimalism.”20 (Daniel Everett)
Ockham’s razor suggests removing theories that add nothing to our understanding, and ignoring claims that cannot be falsified. On that basis, modern linguistics should treat AI as an almost unimaginably fertile playground for testing hypotheses. Chomsky’s own co-authored verdict remains stark:
“The minimal meaning-bearing elements of human languages … are radically different from anything known in animal communication systems. Their origin is entirely obscure.”21 (Berwick and Chomsky)
Yet UG’s core idea is powerful, simple and useful. It echoes Jung’s collective unconscious — inherited instincts, fears and symbolic forms — and focuses attention on our innate, shared capacity for abstraction and symbol. We should stay open, too, to Everett’s counter-claim that little more than general intelligence is innate, and that language emerges from intelligence and the need to build communities.
Chomsky’s most interesting suggestion may be that language did not evolve primarily for communication at all. Its drivers may be aesthetic and emotional as much as functional. Whatever the mix, our universal love of songs, stories and legends suggests that from the beginning we used the creative freedom of language for mirth, myth, mischief and well-meaning lies — or fiction, if you prefer the polite term.
“What I think is a primary ‘fact’ about my work, that it is all of a piece, and fundamentally linguistic in inspiration. The invention of languages is the foundation. The ‘stories’ were made rather to provide a world for the languages than the reverse. To me a name comes first and the story follows.”22 (J.R.R. Tolkien)
Continued in Part 2: The Limits of Logic.
Footnotes #
“As Kingfishers Catch Fire, Dragonflies Draw Flame” (1877). ↩︎
Tractatus Logico-Philosophicus (1922), 4.112. ↩︎
Philosophical Investigations (1953), §109, §255. ↩︎
Terrence W. Deacon, The Symbolic Species: The Co-evolution of Language and the Brain (1997). ↩︎
Alan Turing, “Computing Machinery and Intelligence”, Mind (1950). ↩︎
John Searle, “Minds, Brains, and Programs”, Behavioral and Brain Sciences (1980). ↩︎
John Searle, “Why dualism and materialism fail to account for consciousness” (2010). ↩︎
Douglas Hofstadter, I Am a Strange Loop (2007). ↩︎
Ursula K. Le Guin, A Wizard of Earthsea (1968). ↩︎
Fred Reed, “Promise of AI not so bright”, Washington Times. ↩︎
Wallace Stevens, “Disillusionment of Ten O’Clock” (1915). ↩︎
BioMed Central, “Q&A: What is human language and how did it evolve?” Whether Neanderthals had complex language is contested; recent anatomical and genetic evidence has made the question more open than it appeared a decade ago. ↩︎
Jackie Nordström, “Language as a discrete combinatorial system, rather than a recursive-embedding one”. ↩︎
Charles Darwin, On the Origin of Species (1859; the “breathed into a few forms or into one” wording is from the second edition, 1860). ↩︎
Rebecca Wicker, “Noam Chomsky, Linguistics and AI”; Ron McClamrock, “Marr’s Three Levels” (1991). ↩︎
Cole Porter, “Let’s Do It, Let’s Fall in Love” (1928). ↩︎
Will Sullivan, “Scientists Suggest a New Layer to Crows’ Cognitive Complexity”, Smithsonian. ↩︎
P. V. Voinov et al., “Chimpanzee coordination and potential communication in a two-touchscreen turn-taking game”, Scientific Reports (2020); Jonas Nölle, Stefan Hartmann and Peeter Tinits, “Language evolution research in the year 2020”. ↩︎
Spencer Feingold, “How artificial intelligence is helping us decode animal languages”, World Economic Forum; Anna Peele, “What are animals saying? AI may help decode their languages”. ↩︎
Daniel Everett, “Chomsky, Wolfe and me”, Aeon. ↩︎
Robert C. Berwick and Noam Chomsky, Why Only Us (MIT Press, 2016). ↩︎
Humphrey Carpenter and Christopher Tolkien (eds.), The Letters of J.R.R. Tolkien, Letter 165 (1955). ↩︎


