Janpha Thadphoothon
"The Universe is large enough to accommodate more than one human truth and explanation."
"You may be right, I may be crazy," as Billy Joel famously sang in 1980—and on some mornings, staring into the rapidly changing world of technology, I must admit I feel a little bit of both. I am sure you would agree with me that we are living through one of the most intellectually unsettling moments in human history. Every morning, over a quiet cup of strong black coffee, I find myself reading news reports, academic papers, and philosophical debates that seem to pull the human mind in two completely opposite directions. On one hand, we are told that artificial intelligence has arrived at the threshold of genuine thought—that machines can write poetry, solve complex theorems, pass medical licensing exams, and converse with an eloquence that rivals our brightest students. On the other hand, we are warned just as fiercely that these systems understand nothing at all—that behind the curtain of dazzling fluency lies an empty room, a cold engine of probability shuffling symbols without a spark of awareness.
It has perplexed me for quite some time how two camps of extraordinarily intelligent people—mathematicians, neuroscientists, computer engineers, and philosophers—can look at the exact same phenomenon and reach conclusions that appear irreconcilable. One may ask what "understanding" truly is, and why we argue over it so passionately today. Those were the days when everything was simple, when human beings held an undisputed monopoly on language, reasoning, and creative expression. Like it or not, the world moves on. And as we stand at the crossroads of silicon and spirit, I have come to a realization that brings a strange sense of peace to this intellectual storm: both sides may be right.
The Riddle in the Forest and the Proof in the Machine
As we know, philosophy often begins with a simple riddle that refuses to stay simple. Consider the ancient question that has echoed through classrooms for centuries: If a tree falls in a forest and no one is around to hear it, does it make a sound?
At first glance, it seems like a clever word game. Based on the first impression, any physicist would settle the matter in five seconds. Let's be a bit more scientific. When a massive trunk crashes to the forest floor, it displaces air molecules, sending longitudinal pressure waves rippling outward through the woods. Those acoustic vibrations occur whether a human ear, a deer, or a microphone is nearby to register them. Physics does not wait for an audience. Yet, if you ask a philosopher or a psychologist, they will remind you that sound is not merely vibrating air—it is the conscious, auditory experience created when those waves strike an eardrum and are interpreted by a living brain. Without a listener, there are waves, yes, but there is no heard sound. There is no witness.
Recently, I came across a modern variation of this classic puzzle that cut straight to the heart of our contemporary predicament: If a mathematical proof falls out of a computer, and no human being on earth can understand it, does it make sense?
I think this question captures the existential anxiety of our era better than almost any technical treatise. In 1976, when mathematicians Kenneth Appel and Wolfgang Haken proved the famous Four Color Theorem with the help of a computer that checked thousands of configurations no human could verify by hand in a lifetime, the academic world erupted. Some argue for the triumph of the result: a valid logical deduction is true regardless of the limitations of human working memory. Some argue against calling it a genuine proof at all, insisting that a mathematical proof is not merely a certificate of truth—it is an instrument of human insight, an explanation that illuminates why something must be so.
The renowned mathematical physicist Sir Roger Penrose has long championed this deeper view. Critics such as Penrose would tell you that genuine mathematical understanding is something you simply cannot formulate in terms of mechanical rules. In his view, human insight involves a non-algorithmic leap—a moment of seeing the whole truth in a flash of comprehension that no step-by-step program can replicate. Wisdom from the past hints that truth without a witness is incomplete. Indeed, if we surrender our intellectual agency to machines that produce uncheckable answers, we risk turning science and scholarship into a priesthood of black boxes, where human beings accept conclusions on blind faith rather than genuine illumination.
Inside the Chinese Room
Not only do I agree with the need to examine what happens beneath the surface of machine intelligence, I also believe we cannot have an honest conversation about modern AI without revisiting philosopher John Searle’s famous 1980 thought experiment: the Chinese Room.
You may wish to picture this scenario. Imagine a man who speaks only English—he does not know a single character of Mandarin Chinese, cannot distinguish 天 (sky) from 大 (big), and has no idea how Chinese grammar works. He is locked inside a closed room. Through a small slot in the door, native Chinese speakers outside slip in sheets of paper written with questions in Chinese. Inside the room, the man has a massive, exhaustive rulebook written in English, along with stacks of blank paper and pencils. The rulebook contains instructions such as: "When you see squiggle X followed by squiggle Y, look up Table 42, copy squiggle Z onto a fresh sheet, and pass it back through the slot."
The man works diligently. He matches shapes, follows the instructions to the letter, and slides the resulting characters back under the door. To the native speakers standing outside, the answers coming out of the room are grammatical, culturally nuanced, witty, and completely indistinguishable from those of an educated native speaker. Naturally, the people outside conclude that the person inside the room understands Chinese fluently.
Yet, as a matter of fact, the man inside understands not a single word. He is merely manipulating formal symbols according to their shapes—what philosophers and linguists call syntax—without having the faintest access to their meaning, or semantics.
I notice how eerily this forty-six-year-old thought experiment mirrors the architecture of today’s Large Language Models. What we all know and agree upon is that modern generative AI can produce essays, summaries, translations, and computer code at breathtaking speed. The news has it that these models are passing bar exams and drafting diplomatic cables. According to the media, we are witnessing the birth of artificial general intelligence.
Nevertheless, it is my long-held belief that (though I could be wrong) we must look very carefully at what is actually happening inside the server farms. An LLM does not encounter words as lived experiences. It slices human language into numerical tokens and multiplies matrices across billions of learned parameters to predict which token statistically belongs next. When an AI writes a moving paragraph about grief, autumn rain, or homesickness, it has never felt the chill of November wind, never mourned a lost friend, and never stood on a railway platform watching a train disappear into the distance. It possesses what Searle called derived intentionality—its words carry meaning only because we, the human readers, bring our own lived consciousness to the page and interpret them.
Where the Analogy Breaks Down
Having said that, I realize that dismissing modern AI as "just a lookup table" does not tell the whole story either. I often doubt myself when I lean too heavily toward pure skepticism, because every time I interact deeply with these systems—testing their ability to draw subtle analogies, synthesize cross-cultural concepts, or spot logical flaws in an argument—I sense that something far more sophisticated than a 1980 rulebook is taking place.
I am not an expert, but I have read somewhere that when philosopher Daniel Dennett examined Searle’s Chinese Room, he coined the term "intuition pump" to describe it—a clever story deliberately engineered to steer your gut reaction toward a predetermined conclusion while hiding the complexity of the whole system. Critics of Searle have pointed out for decades that while the man inside the room does not understand Chinese, the entire system—the man, the rulebook, the scratchpads, and the dynamic interactions between them—is what performs the cognitive work.
What's more interesting is that modern neural networks do not operate like static dictionaries or rigid if-then flowcharts. During my own studies in Generative AI with Large Language Models, one concept struck me with particular force: the geometry of vector embeddings. Let me introduce you to the notion of high-dimensional latent space. When a modern neural network is trained on vast corpora of human text, it does not memorize canned responses. Instead, it maps words and concepts into an intricate mathematical universe with thousands of dimensions, where proximity reflects conceptual relationship. In this space, the vector distance between king and queen mirrors the distance between man and woman; clusters of ideas around justice, mercy, law, and ethics organize themselves into structured neighborhoods without a single human programmer writing a rulebook.
I am not sure, but my gut tells me that when a system builds an internal geometric model of how concepts relate to one another across human knowledge, calling it "mere symbol shuffling" misses something vital. Experts say that recent breakthroughs in mechanistic interpretability—where scientists peer inside the hidden layers of neural networks—reveal that these models construct internal representations of spatial maps, timelines, and causal logic. What's more, today's multimodal systems are no longer trapped inside a text-only room. They process images, sounds, video, and robotic sensor data simultaneously, linking the word apple directly to its crimson hue, its crunch, and its physical weight.
People say that a single neuron in the human brain does not understand English either; it merely fires electrochemical signals according to biochemical laws. Yet eighty-six billion neurons wired together give rise to Shakespeare, Newton, and a mother's lullaby. Why, then, should we assume that billions of artificial parameters arranged in complex attention mechanisms are incapable of a genuine form of functional comprehension?
Two Frames of Reference
Gradually, I have come to see why this debate never ends, and why brilliant minds talk past one another at conferences and in journals. Fundamentally, it is all about a collision between two distinct frames of reference—two different ways of looking at the world while using the exact same word: understanding.
I'd like to entertain you with the idea that both sides in this great debate are right, provided we keep their frames of reference separate.
First of all, there is the Third-Person Frame—the functional, relational, and structural perspective. In this frame, we look from the outside in. Understanding is measured by competence, coherence, and the ability to navigate a system of relationships. If a student, a colleague, or an AI system can take an unfamiliar problem, identify its underlying principles, connect it to distant domains, and explain the reasoning clearly, then within that third-person frame, functional understanding has occurred. Here, the AI researchers and functionalist philosophers are right. Meaning, in public communication, is largely built through patterns of use and relational structure.
And then ultimately, there is the First-Person Frame—the phenomenological, subjective, and existential perspective. In this frame, we look from the inside out. Understanding is not merely processing information accurately; it is experiencing what that information means to a living, mortal being. It is the sudden warmth of recognition, the ache of empathy, the moral weight of a choice, and the quiet realization of one's place in the universe. Here, Searle, Penrose, and the humanists are completely right. A machine has no "inside." There is no one home behind the screen feeling the beauty of a theorem or the sorrow of a farewell. Without a conscious subject, there is computation, but there is no witness.
It is well known that confusion arises the moment we collapse these two frames into one. When we mistake third-person fluency for first-person consciousness, we fall into the trap of anthropomorphism—treating software as if it has feelings, intentions, or moral agency. Conversely, when we insist that the absence of first-person consciousness means a machine performs zero genuine cognitive work, we fall into biological reductionism—blinding ourselves to the remarkable analytical power these tools bring to human inquiry.
Reflections from the Language Classroom
As a language teacher, I live inside this tension every single day. For over three decades in university classrooms—from Bangkok to international exchange programs abroad—I have watched students wrestle with the difference between knowing the rules of a language and truly inhabiting it.
In Thailand, for example, many of our students spend twelve years studying English grammar rules, memorizing vocabulary lists, and drilling multiple-choice tests. They become remarkably adept at the mechanics of the Chinese Room. Give them a sentence with a missing preposition or a tricky conditional clause, and many can apply the rulebook with precision. Yet, when they step outside the classroom and face a real human being who asks an unexpected question, they often freeze. Why? Because rules alone are an external scaffold; true communicative competence requires internalizing meaning, reading social context, and daring to express an authentic self.
I must admit that the arrival of generative AI has forced me to rethink how we teach writing, reading, and critical thinking at Dhurakij Pundit University and beyond. Earlier in my career, if a student submitted an essay with spotless grammar, varied vocabulary, and well-structured paragraphs, I naturally assumed that deep thinking had taken place behind those words. Today, that assumption no longer holds. A machine can generate impeccable third-person prose in ten seconds flat. Yet, as the great Italian poet Dante Alighieri once reminded us in his Inferno, 'Consider your seed: you were not made to live like brutes, but to follow virtue and knowledge.' Machines can assemble syntax, but only human beings are called to pursue virtue and genuine knowledge.
In my opinion, this does not spell the end of language teaching; rather, it clarifies our highest purpose. If machines can handle routine syntactic assembly, then human education must double down on the first-person frame—on voice, agency, ethical judgment, and lived reflection. Whenever I work with young writers, whether in formal academic workshops or creative writing clubs, I remind them that writing is not merely producing correct sentences. Writing is an act of witnessing. It is the courage to say, "I stood here, I observed this, I wrestled with this contradiction, and this is what it meant to me."
It's hard to describe, but I will try: there is an unmistakable fingerprint in writing that comes from a lived life. You can hear it in the cadence of a sentence shaped by years of reading, walking, stumbling, and reflecting. An AI can imitate the statistical surface of a writer's style, much like a skilled forger copying brushstrokes, yet it has no stake in the truth of what it writes. It does not risk reputation, heartbreak, or growth.
Living Wisely with Both Truths
It is my personal belief that humanity is entering an era that demands a new kind of intellectual maturity—what we might call cognitive binocular vision. Just as two eyes set slightly apart allow us to perceive depth in the physical world, holding both the third-person and first-person frames in view at the same time gives us the depth perception we need to navigate the age of artificial intelligence.
Globally, we see nations, corporations, and universities rushing either to worship AI as an infallible oracle or to ban it as an existential threat. Both reactions stem from seeing with only one eye open. I like the idea of treating artificial intelligence neither as a rival soul nor as a mere parlor trick, but as an extraordinary mirror and amplifier of human thought.
Fundamentally, I would argue that we need machines precisely because human minds are finite, slow, and prone to fatigue—and machines need us precisely because silicon has no conscience, no mortality, and no capacity for awe. A computer can traverse a labyrinth of ten million logical steps in the blink of an eye to verify a mathematical theorem, yet it takes a human mathematician to look at that result and whisper, "How magnificent." A language model can synthesize centuries of philosophical commentary in seconds, yet it takes a human reader over a morning cup of coffee to feel how those ideas illuminate their own life.
Make no mistake, the questions raised by the falling tree, the uncheckable proof, and the Chinese Room will stay with us for generations to come. No one knows everything, but I would like to suggest that we stop demanding a simple "yes" or "no" to whether machines truly understand. Instead, let us recognize that we are witnessing the birth of a new kind of partner in the long human adventure of inquiry.
My conviction is that as long as we remain clear-eyed about the boundary between functional mastery and conscious meaning, we have nothing to fear from intelligent machines. The computer can verify the steps of the proof, and the model can weave the tapestry of words—but it is still up to us, the living witnesses in the forest, to listen, to wonder, and to decide what it all means.
This brings me back to a quiet conversation I had years ago during my doctoral studies at the University of Canberra in the ACT. One of the professors handed me his book and offered three simple words of advice that have stayed with me ever since: "Open your mind." To this day, I remain deeply grateful to him for that gesture. Looking back now, I realize he was inviting me to resist the comfort of rigid, single-minded explanations and to welcome the vast complexity of human inquiry—much like his great literary ancestor Dante Alighieri, who reminded us in Inferno that we were made "to follow virtue and knowledge," and closed Paradiso with a vision of a universe moved by love and wonder.
But the beautiful thing is that in asking whether machines can understand, we are forced to rediscover what makes human understanding so precious in the first place. The past is the past, and we cannot return to a world before intelligent algorithms. Yet as we walk forward into this uncharted territory, we carry with us that enduring call to keep our minds open—alongside something no rulebook or neural network can ever replicate: the irreducible wonder of being awake to the world. Ultimately, it is my belief that the Universe is large enough to accommodate more than one human truth and explanation, and that should give us hope. I could be wrong, but I somehow think that is more than enough.
About the Author: Janpha Thadphoothon is an Assistant Professor of English Language Teaching at the Faculty of Arts, Dhurakij Pundit University (DPU) in Bangkok, Thailand, and holds a Doctor of Education in Applied Linguistics and TESOL from the University of Canberra. Alongside over three decades of university teaching, he holds a certificate in Generative AI with Large Language Models from DeepLearning.AI. As the founder and director of The Creative English Writing Club of Thailand, his writing and research explore the intersections of human agency, language education, philosophy, and emerging technologies.