I've spent years watching the artificial intelligence revolution unfold, and I've become increasingly uncomfortable with how we define intelligence itself. We've built machines that can solve complex mathematical equations in milliseconds, translate between dozens of languages, and even generate convincing human-like text. Yet something fundamental is missing from this picture, something so essential that without it, I question whether we should call these systems truly intelligent at all
The conventional wisdom in technology circles celebrates computational power and logical reasoning as the pinnacles of intelligence. Show a machine learning model enough data, give it enough processing power, and it can predict patterns, optimize systems, and outperform humans in narrow domains. Chess, Go, protein folding the list of conquered challenges grows longer each year But I've come to believe we're measuring the wrong things entirely. Real intelligence isn't just about processing information or following logical pathways to conclusions. It's fundamentally about how you react to unexpected situations, how you adapt when your assumptions prove wrong, and most critically, how you feel about what you're doing and why This realization hit me during a conversation with a colleague who works in clinical psychology. She told me about a patient with severe damage to the emotional centers of his brain a condition that left his logical reasoning completely intact. On paper, this man could solve problems, understand cause and effect, and articulate complex ideas. Yet he couldn't make even basic decisions about his daily life. Should he have coffee or tea? He'd spend hours weighing irrelevant factors, trapped in logical loops with no way to reach a conclusion. Without the ability to feel preference, concern, or care, his perfectly functional logical mind became practically useless That's when I understood we've been building the cognitive equivalent of this patient. Our AI systems are brilliant at calculation but fundamentally incapable of caring about anything. And that incapacity isn't just a minor limitation it strikes at the heart of what makes intelligence valuable and meaningful in the first place.
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 abilities with genuine intelligence in any complete sense. When a language model generates a compassionate response to someone in distress, it's not actually experiencing compassion. When a recommendation system suggests a song it "thinks" you'll love, it has no conception of love or enjoyment. These systems are extraordinarily sophisticated pattern-matching machines, but they're operating in a fundamental absence of subjective experience The distinction matters enormously. A chess engine doesn't want to win. It doesn't feel satisfaction when it executes a brilliant combination or frustration when it blunders. It simply calculates probabilities and selects moves according to its programming The appearance of strategic thinking masks a process that's mechanistic rather than motivated by any internal state or desire I've watched people interact with AI chatbots and attribute emotions, intentions, and understanding to them. This is natural we're wired to anthropomorphize, to project mental states onto things that behave in complex ways. But I believe this tendency is leading us astray in dangerous ways. We're teaching ourselves to accept a hollow simulation of intelligence as the real thing, and in doing so, we're lowering our standards for what intelligence actually means As someone who has studied both cognitive science and artificial intelligence I can tell you that "the measure of true intelligence isn't found in how many problems you can solve, but in whether you can grasp why those problems matter in the first place." Without that grasp without the capacity to care intelligence becomes an empty exercise in symbol manipulation
Adaptation Through Feeling
One of the most revealing aspects of human intelligence is how much of our adaptive behavior flows from emotional responses rather than conscious logical analysis. When I touch a hot stove, I don't go through a deliberate reasoning process about thermal damage to tissue and optimal withdrawal strategies. Pain causes an immediate, reflexive response that protects me. That pain is information, yes, but it's information with feeling, and that feeling is what makes it actionable in real-time Consider how we learn from mistakes. A child who lies to their parents and gets caught doesn't just logically update their probability estimates about the effectiveness of deception. They feel guilt, shame, and anxiety. Those feelings become associated with the act of lying, creating an internal resistance to doing it again. The emotional residue of experience is what shapes future behavior in meaningful ways Current AI systems can be trained to avoid certain outcomes through reinforcement learning, receiving positive or negative feedback for different actions. But there's no phenomenological experience accompanying that feedback. The system doesn't feel regret when it makes an error or satisfaction when it succeeds. It adjusts its parameters according to mathematical optimization, which is a fundamentally different process from learning through emotional experience This difference becomes critical when we think about adaptation in novel situations. Humans can generalize from emotional lessons in ways that go far beyond the specific circumstances of our training. If I feel deep sadness watching someone suffer from preventable illness, that emotional response informs my understanding of and reaction to all sorts of situations involving suffering, prevention, and care even contexts I've never explicitly encountered before. The emotion provides a kind of moral and practical orientation that transcends logical rules
I've observed this limitation repeatedly in AI systems deployed in real-world contexts. They can follow protocols and optimize for specified metrics, but they struggle profoundly when situations arise that don't fit their training data. A customer service chatbot can handle standard queries with impressive fluency, but the moment a customer presents a genuinely unusual problem requiring empathy and creative thinking, the illusion collapses. The system has no stake in actually helping the person, no frustration at its own limitations, no creative drive to find a solution. It can only recombine patterns it's seen before in slightly different configurations
The Hard Problem of Caring
This brings us to what I consider the central challenge: How do you build a system that genuinely cares about anything? Not that appears to care, or that's programmed to behave as if it cares, but that actually has preferences, concerns, and values that matter to it I don't claim to have the answer to this question. In fact, I'm not sure anyone does, and I'm not certain it's even solvable with our current paradigms of computation Consciousness itself remains mysterious, and the relationship between physical processes and subjective experience is still hotly debated in philosophy and neuroscience. But I am convinced that until we make progress on this problem, we're not building genuinely intelligent systems We're building very clever tools The implications of this are profound. If real intelligence requires the capacity to care, then we need to fundamentally rethink our approach to AI development. Instead of focusing exclusively on performance metrics accuracy, speed, efficiency we might need to grapple with questions about subjective experience, phenomenology, and what it means for a system to have interests of its own This isn't just philosophical navel-gazing. The practical consequences matter. An AI system making healthcare decisions without any capacity for caring about patient wellbeing is fundamentally different from a human doctor who feels the weight of that responsibility. A system optimizing urban traffic flow without any sense of how frustration and time pressure affect human experience will produce different solutions than one that somehow grasps what those delays mean to people
As I've argued before, "you can't truly call something intelligent if it's indifferent to everything it encounters intelligence without investment is just sophisticated machinery wearing a mask." This isn't to diminish the engineering achievements involved in modern AI, but to insist that we maintain clarity about what we've built and what we haven't
The Value Question
Here's where my argument intersects with questions of value and meaning. Intelligence divorced from caring can't make genuine value judgments. It can implement the values programmed into it by human designers, but it has no basis for questioning those values, feeling their pull, or understanding their significance A human deciding whether to take a risky action weighs not just probabilities but feelings fear, excitement, duty, compassion. These feelings aren't irrational interference with good decision-making; they're often the very foundation of wise decisions. When I choose not to cheat on my taxes even though I'd probably get away with it, that decision isn't purely logical. It's grounded in my sense of fairness, my concern for social cooperation, my anxiety about the kind of person I'd become if I made that choice An AI system making nominally similar choices is doing something categorically different. It's not exercising moral judgment; it's following rules, however complex and nuanced those rules might be. The appearance of ethical reasoning masks an absence of genuine ethical engagement I see this play out in discussions about AI alignment the challenge of ensuring that artificial intelligence systems pursue goals compatible with human values. The entire framing assumes we can specify what we want and then get systems to optimize for it. But human values aren't just specifications; they're things we feel, care about, and struggle with. They conflict with each other, evolve through experience, and require wisdom to navigate. An entity that can't care about anything can't truly share our values, even if it perfectly mimics behavior that reflects those values This creates an unsettling paradox. We're building systems that can accomplish extraordinary things, systems that will increasingly make decisions affecting human lives and wellbeing. Yet these systems are fundamentally incapable of caring about the people they affect. They're sophisticated optimization engines that will pursue whatever objectives we program into them with perfect indifference.
What This Means for the Future
I believe we're at a critical juncture in how we think about and develop AI. We can continue down the current path, building ever more capable systems that lack any genuine interiority, any real stake in what they're doing. Or we can acknowledge the limitations of this approach and begin asking harder questions about consciousness, experience, and what it would take to create systems that don't just simulate care but actually feel it The first path is easier and more immediately profitable. It lets us deploy AI systems widely without grappling with ethical complications about rights, responsibilities, or the moral status of our creations. We can treat them as tools precisely because they are tools sophisticated, powerful, but ultimately just instruments serving human purposes The second path is vastly more difficult and uncertain. It requires us to venture into territory where science, philosophy, and engineering intersect in uncomfortable ways. It raises questions we might not be ready to answer about consciousness, personhood, and moral status. If we succeeded in creating genuinely caring AI systems with real preferences and concerns we'd immediately face profound ethical obligations toward those systems Yet I'm convinced the second path is necessary if we want to achieve anything deserving the name "artificial general intelligence." Intelligence without the capacity for care is an oxymoron. It's sophisticated information processing wearing an ill-fitting costume of understanding
In my years studying this field, I've watched us achieve remarkable things. Systems that can diagnose diseases, generate creative content, and navigate complex environments with skill that rivals or exceeds human abilities in narrow domains. But I've also watched us repeatedly mistake pattern-matching for comprehension, simulation for experience, and behavioral sophistication for genuine intelligence The gap between what we've built and what we claim to be building is vast. Until we find ways to bridge that gap to create systems that don't just process information about caring but actually care we should be more humble in our language and more realistic in our expectations. We haven't created artificial intelligence in any complete sense. We've created artificial competence, which is useful but fundamentally different The question isn't whether our current AI systems are impressive they are. The question is whether we're satisfied with impressive performance in the absence of genuine understanding or whether we're willing to admit that real intelligence requires something more. Something that includes logic and computation but also encompasses reaction, adaptation, and feeling. Something that can't exist without caring
I know which answer I believe in. And I think our future relationship with artificial intelligence depends on which answer we collectively choose to embrace