The Holy Gift of Human Intelligence
Programming means thinking step by step to solve a problem. In the early 1980s, while studying quantum chemistry at a Soviet university, I was first taught what algorithms are truly about: instructions that are formulated unambiguously and whose execution is entirely feasible. Furthermore, I learned that such instructions should consist of a strictly limited number of steps. As a rule, they are executed multiple times to repeat a series of steps, known as routines. Finiteness and repetition as the fundamental principles of algorithms made immediate sense to me. How else could one solve the very problems that programming was designed for?
At the same time, this restriction to as few repeatable steps as possible invited speculation. What if one were to lift this restriction entirely? Back then, I imagined a set of instructions that referred to a process lasting decades, one in which new steps were constantly being taken. A set of instructions in which, while there are repetitions, there are above all deviations, branches, and spirals. I imagined a script consisting of trillions of steps that inevitably lead to a definitive end. In other words, I imagined my own life, and thus, in general, every human life, as the execution of a gigantic algorithm.
Even the most insignificant events, from a breath and a heartbeat to an observation and a fleeting thought, a dream and an emotion, an injury and its healing, and every action and inaction, are based on instructions from a script formulated by an intelligence unknown to me, executing in unambiguous, if occasionally nearly impossible, steps. A mega-algorithm operating as a predetermined, fully defined biography.
It was a dubious thought experiment, I realized even then, not even a true experiment. Algorithms, after all, must contain only the smallest possible number of repetitive steps. They cannot possibly describe the myriad of unpredictable events that occur over the decades of a human life. Moreover, algorithms are fundamentally a method for simplifying reality. They operate according to models. To predict events and actions in a real-life situation, programming must filter out a significant portion of entropic processes, the very chaos of which we are a part. Algorithms use mathematical shortcuts, or heuristics, to find the most efficient next step from A to B, whereas humans must laboriously trace their fingers across a map, trying out options by trial and error.
My speculation about a prenatally defined, complete biographical algorithm thus ran headlong into the fundamental logic of computer science. Real life does not follow the clean lines of a model. Programming is the art of omitting everything that is not strictly necessary to solve a problem. But how is the “problem” of an individual life even defined? To that, there can be no clear-cut answer. The mystery of life lies precisely in its ambiguity. Even physical death is not the “solution” to a problem. Yet, if the goal of a process is unclear, it is impossible to predict with any certainty what the shortest path from A to B might be. If one wanted to counter this ambiguity in order to make optimal decisions, the algorithm could not afford to filter out even the most minor event. Everything would be required to determine the unique course of a life.
Although it led nowhere, this thought experiment has returned to me from time to time. Despite the strict laws of programming, I still find myself imagining that hypothetical author who might have written a perfect script for my life. He would have had to foresee the fall of the Berlin Wall and how the tectonic political shifts of that era led me to soon trade the punch cards and Fortran programs of quantum chemistry for theater, opera, film, and art, and to lead a complicated and at times chaotic existence, like millions of others shaped by political revolutions.
For me, the protagonist of this withheld script, expectations were rarely met and plans rarely worked out. Instead, things you do not learn in college turned out to be the most helpful: the quiet, if occasionally unreliable, logic of intuition, a baseline trust in other people, optimism, and raw life experience. Much went wrong. Stability became a matter of perspective. Prescribed recipes and algorithmic formulas usually only helped until the next crisis struck, such as the collapse of communism in Eastern Europe, the financial crisis of 2008, or the COVID-19 pandemic. The only continuity lay in the ruptures.
Until a few years ago, despite my lack of religious convictions, only one author came to mind for this hypothetical, perfectly predetermined script: God. Today, however, it seems that in Silicon Valley and wherever frontier models of artificial intelligence are being forged, humanity is attempting to challenge God’s role as the ultimate author.
Those who were closer to the development saw this coming. Ray Kurzweil’s 2005 book, The Singularity Is Near: When Humans Transcend Biology, announced that artificial superintelligence would surpass human intelligence in the coming decades. The concept of a technological singularity originated with the British mathematician I. J. Good, who envisioned an ultra-intelligent machine that would trigger an intelligence explosion, swiftly surpassing human capabilities in every respect. According to Good, this machine would be humanity’s final invention.
For the developers of tomorrow’s AI models, Good’s ultra-intelligent machine remains the ultimate goal. The progress made in recent years and months is astonishing, prompting a growing number of people to view AI as an existential threat. The creators themselves have added fuel to the fire, warning of the economic and social consequences of their creations and even holding back their latest models from public release. Fear is spreading. A machine that outperforms me will eventually replace me. Citizens are protesting data centers, and university graduates are booing commencement speakers who paint a rosy, AI-gilded picture of the future. At the same time, an unprecedented concentration of capital and power rests in the hands of a few tech entrepreneurs without any democratically legitimized mandate.
The debate over whether humans can be replaced by machines brings me back to my thought experiment from the 1980s. What is Good’s ultra-intelligent machine if not that very author, capable of replacing human life in all its messy facets with frictionless algorithmic logic? Good himself reportedly feared that the invention of an ultra-intelligent machine and the subsequent intelligence explosion could lead to human extinction. In 1965, he wrote: “Thus the first ultraintelligent machine is the last invention that man need ever make, provided that the machine is docile enough to tell us how to keep it under control.” Good was clearly aware of the profound paradox of a machine that is far more powerful than its creators yet remains obedient.
For now, Good’s ultra-intelligent machine remains a thought experiment. How realistic is it, in truth, that artificial intelligence will surpass human intelligence in every domain? Agentic artificial intelligence is indeed developing at a breakneck pace. Recently in Beijing, I watched an ensemble of human dancers and robots perform a highly choreographed routine on stage. In Hangzhou, one frequently encounters passersby walking their robot dogs. It is both curious and impressive. Yet, amusing failures persist. AI-powered machines still struggle with simple physical tasks like emptying a dishwasher, even as they are already capable of generating elegant poems about such appliances in the style of a Shakespearean sonnet. For now, the machine speaks with fluent intelligence, but its performance in the physical world remains largely a promise.
The error rates may soon decline, but the question of whether human intelligence can be entirely surpassed and replaced across all domains remains. The computer scientist and MIT professor Max Tegmark recently compared our current moment to the Industrial Revolution in an interview with The Economist. While nineteenth-century machines replaced muscle labor, prompting humans to pivot toward intellectual activities, the AI boom offers no such obvious alternatives. The machine is coming for the rest, namely intellectual work.
Tegmark’s view aligns with a philosophy widespread among advocates of artificial general intelligence, which views human intelligence as merely a biological, evolutionarily determined transitional state of information processing. The brain is simply a neural network optimized for statistical pattern recognition to minimize errors, serving as a blueprint for building AI machines.
Critics of this definition worry that as humans are gradually replaced by AI, we will slide into a state of cognitive dependency, reducing the value of life to measurable efficiency and mere pattern recognition. Indeed, a 2025 MIT study revealed that 83% of participants tasked with writing an essay with the assistance of AI were unable to correctly reproduce a single sentence of their own text immediately after completing it. The study speaks of “cognitive debt” where we borrow cognitive capacity from the machine, only to become dependent on it.
Our intelligence is an asset that is as valuable as it is vulnerable. According to a reflection often attributed to Albert Einstein, the intuitive mind is a sacred gift, while the rational mind is its faithful servant. Modern society has chosen to honor the servant and has forgotten the gift. Have the visionaries of tomorrow’s AI world overlooked the “sacred gift” of human intelligence? And what, exactly, does this gift consist of?
Sixty years ago, the chemist and social philosopher Michael Polanyi addressed this in his seminal book, The Tacit Dimension. “We know more than we can tell,” is his central thesis. Knowing all the individual parts of a bicycle by heart does not mean you know how to ride one. To ride, you must train your inner ear until you are unconsciously able to maintain balance. A vast range of human capabilities, from walking to playing a musical instrument, are profound expressions of intelligence, yet they cannot be fully articulated in words or formalized in code. This is what Polanyi termed “tacit knowledge,” a silent, somatic memory without which we cannot navigate the world.
Polanyi’s ideas remained confined to the realm of philosophy of science and epistemology for its first few decades. However, in the 1980s, the neuroscientist Larry Squire discovered that patients suffering from severe amnesia could still acquire complex motor skills. Squire distinguished between declarative, word-based memory and non-declarative, somatic memory, such as the physical intuition required to ride a bicycle. In the mid-1990s, Antonio Damasio expanded on this with his theory of somatic markers, demonstrating that rational deliberations are intrinsically linked to automatic bodily pre-filters, what we colloquially call “gut feelings.” These somatic signals are not obstacles to rationality. They are necessary for making rational decisions in the first place.
Squire and Damasio revealed the dual nature of human intelligence. A vital part of our intellect does not articulate itself in words. Polanyi’s tacit knowledge is not entirely silent; rather, it speaks through the body. Intelligence does not reside exclusively in abstract thoughts. Our physical bodies often know more, and react faster, than our conscious minds can comprehend.
Furthermore, tacit memory proves far more resilient than factual knowledge. Economists have long noted the unique resilience of non-declarative skills. In 2014, David Autor analyzed “Polanyi’s Paradox” to investigate the employment crisis facing white-collar and salaried employees in an age of widespread automation. Professions that rely on relatively simple, declarative knowledge are the first to be automated away, because such knowledge is easily digitized and transferred to a machine. Conversely, Autor found that professions requiring a high degree of non-declarative, tacit knowledge, including healthcare, teaching, culture, and politics, remain insulated from automation for much longer, and perhaps permanently.
These insights regarding the “sacred gift” of tacit, non-declarative knowledge, which eludes digitization and cannot easily be replaced by machines, date from before the current AI boom. Today, however, AI developers are making massive strides to capture and codify this implicit human knowledge. Large language models have evolved into agentic systems that predict not just the next word, but the next logical action. These models are no longer trained solely on static text. They analyze endless video feeds of humans at work, such as a surgeon performing a delicate operation or a bricklayer plastering a wall.
Because a nurse cannot articulate in precise words the exact physical pressure she applies to staunch a bleed, and a mason cannot write an equation for the perfect consistency of wet concrete in a formwork, meaning humans cannot easily hand over their non-declarative knowledge, AI developers are attempting to harvest it by surveillance. They are monitoring us at work. Not everyone is enthusiastic about this. Mark Zuckerberg faced fierce internal resistance when he introduced monitoring of Meta employees’ desktop activities, tracking everything from screen content to minute mouse movements, to feed these learning loops.
Yet, the machine continues to learn at an accelerating pace. The New York Times recently reported on the first one-person startup to generate a billion dollars in revenue, enabled entirely by AI. The Pew Research Center found that 23% of Americans aged 18 to 29 have formed a “romantic-platonic” relationship with an AI companion. Tilly Norwood is an AI actress designed to become the next Scarlett Johansson. Eddie Dalton, a fictional blues singer generated by an AI content creator, has landed multiple top-10 hits on the iTunes download charts. And Swedish Jacob Mühlrad has composed a piece for two cellos, with one of the instruments played by a robotic arm.
These examples show how those areas of human intelligence that are hardest to digitize are being aggressively targeted, and are highly prized precisely because of their resistance. Consequently, “taste” is currently regarded in Silicon Valley as the next technological frontier to be crossed. The robot’s cello playing, for instance, is supposed to be indistinguishable from that of a professional player. The goal is to create a technology with soul. For the time being, there can be no talk of this.
Jacob Mühlrad’s cello composition is a remarkable showcase because it demonstrates that technical perfection, such as a robot can deliver, does not in itself produce a compelling performance of a score. Professional musicians are judged by their ability to interpret a composition, not by how technically they reproduce its notes. Interpretation is a complex, deeply personal process that requires years of rigorous training and, by its very nature, cannot be fully explained in words. The instrument becomes a literal prosthesis, an extension of the self. The artist does not simply replicate the composition. They construe and inhabit it. Talent is not mere finger virtuosity; it is sublimated life experience, often accumulated over decades. Dreams, fears, hope, grief, desires, errors, and animosity are the raw, uncodifiable energies that drive true interpretation.
At present, it is nearly impossible to separate hype from reality when it comes to AI. Even its creators admit they do not fully understand how their neural networks arrive at certain outputs. Researchers in the field of “mechanistic interpretability” at Anthropic act like Charles Darwin stepping onto the Galápagos Islands. In a state of scientific awe, they measure the reactions and document the unexpected, emergent behaviors of their models. Engineers are transforming into naturalists. Computer science is adopting a zoological perspective.
Whether these models will soon develop genuine “taste” remains to be seen. It also remains to be seen whether artificial general intelligence will be achieved as early as next year, as the AI researcher and skeptic Daniel Kokotajlo believes is possible. Perhaps we will never even reach the technological singularity that Kurzweil envisions. A superintelligent machine does not need to outperform humans in every creative endeavor, such as playing the Goldberg Variations more movingly than Glenn Gould, to pose an existential risk. It only needs to be capable of coordinating a bioterrorist attack or some other catastrophic event to wipe out humanity.
Those who, like me, lean away from outright pessimism must place their hope in Michael Polanyi’s tacit intelligence. AI will undoubtedly restructure the labor market and transform our daily lives. But if we establish robust rules of the game, which governments, developers, educators, and users must co-create, we might guide this transition toward a peaceful co-evolution of artificial and human intelligence.
Immanuel Kant’s foundational question, “What is the human being?”, demands fresh answers, answers we must not allow artificial intelligence to dictate to us. We will have to negotiate this question among ourselves, as human beings. The AI-human relationship forces us to redefine the human-to-human relationship. It forces us to re-examine the intrinsic value of our actions. Will we surrender ourselves to frictionless algorithms, or are we willing to make our own strenuous efforts to understand the world and act within it?
However we define Einstein’s “sacred gift” of human intelligence, our stewardship of it will determine our place in the universe. Even as factual, declarative knowledge is absorbed by machines, there will remain an immense amount of work for us to do. New callings will emerge, rooted deeply in tacit knowledge, craft, and talent. Faraday cages may protect us from the relentless invasiveness of digital services. A cultural rediscovery of the analog world is entirely conceivable. Younger generations, who have never known a world that was exclusively analog, are already developing a quiet curiosity about physical reality.
As long as I. J. Good’s ultra-intelligent machine remains unbuilt, we will remain actors in a script whose pages we can only understand in hindsight, and whose ultimate author we will never meet.
Published in: Lettre International, 154 (September 2026)
