I've been thinking about this paradox for months now, and it keeps me up at night. Every time artificial intelligence achieves something we once thought impossible beating world champions at chess, diagnosing diseases, writing poetry, generating photorealistic images we move the goalposts. We say Well, that's not real intelligence. Real intelligence is something else entirely And I'm starting to wonder are we right to do this, or are we simply protecting our egos The question becomes even more urgent when we consider what might happen next. We've always said that strong AI would be able to Find solutions mining through vast datasets, identifying patterns, connecting dots that humans miss. And it's true that perhaps 90% of what current AI systems do involves pattern recognition from existing, discovered knowledge they've been trained on. But what happens when AI reaches the stage where it doesn't just find solutions but creates them? Will we change the definition of intelligence yet again
To answer this complex question, I believe we first need to address a more fundamental one: what are the chances that we'll actually reach a stage where AI creates solutions rather than merely searching for them? And here's where I think we run into a massive, perhaps insurmountable problem one that goes to the very heart of what intelligence really means
The Crisis of Care
What truly makes humans intelligent isn't just our computational power or our ability to process information. It's our capacity to care about creating solutions. It's our curiosity, our obsession, our late nights spent wrestling with problems not because we're programmed to but because we're genuinely interested in the answers. AI, for all its impressive capabilities, doesn't care about anything at all
Let me put this more bluntly We've been building the cognitive equivalent of a brain-damaged patient who has lost all sense of feeling. Our AI systems are brilliant at calculation, but they are fundamentally incapable of caring about anything. And this incapacity isn't just a minor limitation it's at the very core of what makes intelligence valuable and meaningful in the first place This realization hit me hard when I was working with a state-of-the-art language model last year I asked it to help solve a thorny ethical dilemma I was facing in my work. The response was sophisticated, nuanced, and completely useless. Why? Because the system had no stake in the outcome. It didn't care whether I made the right choice or ruined someone's life. It was simply pattern-matching against billions of tokens of training data, producing statistically likely sequences of words that resembled human reasoning without any of the underlying motivation that makes reasoning matter
Think about the greatest innovations in human history. They didn't come from pure computational power They came from people who were obsessed with problems, who couldn't let go of questions that haunted them. Marie Curie worked with radioactive materials that eventually killed her because she cared deeply about understanding atomic structure The Wright brothers risked their lives repeatedly because they were fascinated by flight. Tim Berners-Lee invented the World Wide Web not because an algorithm told him to but because he was frustrated by the difficulty of sharing information between researchers
Can an AI system ever have this kind of drive I'm increasingly convinced the answer is no not with our current paradigms, anyway And this matters enormously when we talk about AI creating rather than finding solutions.
The Illusion of Understanding
Let me be clear about what I'm arguing here. I'm not saying that current AI systems aren't impressive or useful they absolutely are. What I'm saying is that we shouldn't confuse their capabilities with genuine intelligence in its fullest sense. When a language model generates an empathetic response to someone in distress, it's not displaying real compassion. When a recommendation system suggests a song it "thinks" you'll like, it has no concept whatsoever of liking or enjoyment. These systems are extraordinarily sophisticated pattern-matching machines, but they operate in the complete absence of subjective experience This distinction is crucial A chess engine doesn't want to win. It doesn't feel satisfaction when it executes a brilliant combination, nor frustration when it makes a blunder. It simply calculates probabilities and selects moves according to its programming. The appearance of strategic thinking masks a mechanical process driven by no internal state or desire whatsoever
The greatest trick modern AI has pulled is making us believe that simulation equals experience. But a perfect simulation of hunger doesn't feel hungry, and a flawless model of curiosity isn't curious. We've mistaken the map for the territory, and in doing so, we've fundamentally misunderstood what we're building
I experienced this disconnect viscerally during a project where I was using AI to help design educational curricula. The system could generate learning objectives, suggest activities, and even predict which students might struggle with certain concepts. But when I asked it to explain why education matters, why we should care about whether students truly understand versus merely memorize, it could only regurgitate philosophical arguments it had seen before. There was no fire behind the words, no genuine conviction. It was like asking a book to care about its contents This brings us back to the question of creation versus discovery. Human creators aren't just recombining existing elements though that's certainly part of the process. They're driven by dissatisfaction with the status quo, by a vision of something better, by an itch they can't scratch. Picasso didn't paint "Guernica" because an algorithm suggested it would be a statistically optimal artwork. He painted it because he was horrified by the bombing of a Spanish town during the Civil War and needed to express that horror.
The Innovation Paradox
Here's where things get really interesting and troubling If AI can't truly care about problems, can it genuinely innovate? Or will it always be limited to extremely sophisticated interpolation between existing ideas
I've seen AI systems produce what looks like creative work. I've watched them generate novel molecular structures for potential drugs, design architectural plans for buildings that have never been built, and compose music that moves people to tears. But when I dig deeper, I always find the same thing: these systems are working within possibility spaces defined by their training data. They're finding unexpected combinations, sure, but they're not asking the fundamental questions that drive paradigm shifts Consider how scientific revolutions actually happen. They don't typically come from incremental improvements to existing theories. They come from people who are willing to question basic assumptions, who notice anomalies that everyone else ignores because those anomalies are interesting to them. Einstein didn't develop relativity by processing more data about Newtonian mechanics. He developed it because he was puzzled by thought experiments about riding alongside a beam of light puzzles that fascinated him personally.
Can we program fascination? Can we create artificial curiosity that's genuine rather than simulated? I'm deeply skeptical. The AI systems I work with can be programmed to explore widely, to try unexpected combinations, to optimize for novelty. But that's not the same as wondering about something, as being kept awake at night by a question you can't answer This matters enormously for the future of innovation. If AI can only work within conceptual frameworks established by humans, then we're not creating artificial intelligence so much as artificial expertise systems that can operate at superhuman levels within existing paradigms but can't establish new ones. That's still incredibly valuable, but it's not what we've been promising. It's not artificial general intelligence; it's artificial specialized intelligence, no matter how many specializations we manage to combine.
The Moving Target of Intelligence
So we return to the original observation: every time AI gets smarter, we change what we mean by intelligence. And I think I understand why we do this now. It's not just defensiveness or human chauvinism, though there's probably some of that involved. It's that we're gradually realizing that intelligence isn't a single scalar quantity that you can have more or less of. It's a constellation of capacities, and some of the ones we thought were most fundamental like caring, curiosity, and genuine understanding turn out to be the hardest to reproduce When AI beats us at chess, we say "that's just brute force calculation, not real intelligence." When it translates languages, we say "that's just pattern matching When it generates art, we say "that's just recombination of existing styles." And there's truth to all these dismissals. But I think we're also revealing something about what we actually value in intelligence: not just the ability to process information and produce outputs, but the capacity for genuine understanding, for caring about truth, for being motivated by curiosity rather than optimization functions
This is why I'm skeptical that we'll see AI truly creating rather than finding solutions anytime soon not unless we solve the problem of artificial motivation, of making systems that genuinely care about the problems they're solving. And I have no idea how we'd do that. I'm not even sure it's possible without creating something that would have moral status, that would deserve rights and considerations, that would be conscious in a way that makes our current use of AI systems deeply problematic
Where This Leaves Us
I don't want to end on a note of despair, because I think AI is going to be enormously valuable regardless of whether it achieves what we might call "true" intelligence. Systems that can find solutions within existing frameworks, that can process vast amounts of information and identify patterns humans would never see, that can automate complex tasks and free humans to focus on higher-level concerns these are all incredibly powerful capabilities.
But I do think we need to be honest about what we're building and what we're not. We're not on the verge of creating artificial minds that will care about solving humanity's problems. We're creating extremely powerful tools that humans will need to wield wisely. The responsibility for caring, for deciding what problems are worth solving, for ensuring that solutions serve human flourishing rather than mere optimization all of that remains with us And maybe that's how it should be. Maybe the capacity to care, to be genuinely curious, to find meaning in the pursuit of understanding maybe these are the things that make biological intelligence special, valuable, worth preserving even as we build ever more capable artificial systems alongside it.
The question isn't whether AI will become intelligent by our standards. The question is whether we'll recognize that intelligence comes in different forms, serves different purposes, and that the kind of intelligence worth having might be the kind that can't be programmed or simulated only lived I started this piece wondering whether we change the definition of intelligence out of defensiveness. I'm ending it convinced that we change the definition because we're slowly learning what intelligence actually is. And what we're learning is that it's stranger, more precious, and more deeply tied to caring and consciousness than we ever imagined. That's not a comfortable realization in an age of rapidly advancing AI, but it might be a necessary one The real test isn't whether AI can create solutions rather than find them. The real test is whether we can maintain our own curiosity, our own capacity to care, our own drive to understand not in competition with our artificial tools, but in collaboration with them. Because at the end of the day, only something that can truly care can decide what's worth creating in the first place.