Learning Without Understanding
If there’s one thing that advanced language models (LLMs), the cutting-edge artificial intelligence technology at the heart of many of today’s AI discussions, have taught us, it’s that learning can occur without understanding.
Neural Networks, Implicit and Human Learning
Advanced language models, built on neural networks, don’t store information in the way we might expect from a human mind. Instead, they “digest” vast amounts of data, transforming it into a complex web of connections that represent recurring patterns in the data itself. Once trained, this network is no longer interpretable in a linear fashion: from a model trained to recognize pictures of cats, for example, we can’t extract a definition of a “cat” or a set of explicit rules for recognizing one in a new image. The model does not ‘understand’ what a cat is, but learned to identify them through a process that remains for us a black box.
What may seem strange about how advanced language models “learn without understanding” is, in fact, a dynamic that’s very common in human learning. We learn a wide range of skills and behaviors without ever understanding the underlying theoretical principles. We learn to ride a bike without knowing anything about the angular momentum that keeps us balanced. We instinctively avoid touching hot objects without having any knowledge of entropy or thermodynamics. We automatically calculate distances with our two eyes, using depth perception, without needing to know the trigonometric formulas that describe the process. When we catch a ball in mid-air, our brain performs complex ballistic predictions, but we don’t need to know the differential equations governing the motion to succeed.
In all these cases, as in neural networks, there is no explicit “knowing,” but a kind of learning based on experience, observation, and practice. It’s a type of knowledge that allows us to act effectively without deeply understanding the physical or mathematical laws underlying those actions.
And if the mentioned abilities are something that more or less we share with animals, implicit learning plays a fundamental role even in more advanced contexts, like school or professional environments. A student who completes many math exercises is generally more skilled than one who has only studied the theory, and the same goes for someone who reads a lot compared to someone who just knows grammar perfectly. A professional with years of experience is almost always more effective than a recent graduate, despite the latter having more up-to-date and fresh knowledge in memory.
Get Vincenzo Piombo’s stories in your inbox
Join Medium for free to get updates from this writer.
In all these cases, it’s implicit learning that makes the difference: the student who has solved many problems wouldn’t be able to explain exactly what he has learned more than the one who focused only on theory. Similarly, the experienced professional might not always rationalize why he is more efficient. It’s the practical exposure, the lived and repeated experience, that refines skills we can’t fully describe, but that makes our actions more effective.
Consciousness: Between the Narratable and the Inaccessible
Consciousness, as I perceive it, seems to have many faces. There’s the one we all tell ourselves and others, the one that allows us to present ourselves to the world as a coherent ‘I,’ in line with social expectations. Then there’s a more secret consciousness, the one that we entrust only to those who are really close to us or perhaps to a psychoanalyst. Yet, both versions remain within the boundaries of what can be narrated; they are aspects of us that we know how to describe and explain.
But then there’s another level of consciousness, one that I feel escapes language. Not because it’s necessarily something to hide, but because words fail to grasp it, as if it belongs to a realm that language cannot map. Perhaps it’s what Jung called the ‘Shadow,’ or maybe it’s just a part of ourselves that defies categorization. It’s there, it’s real, but it’s not narratable. It can be somehow reached through meditation or particular emotional conditions.
When I reflect on how we come to know ourselves, the most authentic part of us resides precisely in that inaccessible zone. Most of our decisions are really taken there. We can’t explain it, but we feel it. Ultimately, it’s like learning who we are without really understanding it, much like a neural network that recognizes something without being able to describe it. We live with ourselves without necessarily having to fully understand ourselves.
Spirituality and Artificial Intelligence
When I think about religion, I’m struck by how the morality it presents seems always directed at the “narratable” part of ourselves. Rules, precepts, and ethical prescriptions are all aimed at an ‘I’ that we can explain, an ‘I’ that acts according to clear and shared principles. And yet, this morality often feels distant from the deep, non-narratable complexity we feel within, where good and evil, if they exist at all, are incredibly difficult to untangle.
Even the idea of God, as formulated by religions, seems aimed at that part of our consciousness that requires definitions and explanations. But every attempt to enclose God in human language, within a comprehensible universe, inevitably runs into the concept of dogma. It’s as if we need to set clear boundaries, to give a comprehensible shape to something that ultimately escapes us.
But if we turn our gaze toward that part of ourselves that is not narratable, we might be able to intuit another kind of spirituality, one that doesn’t require dogma. We don’t understand God any more than we fully understand ourselves, but we can “learn” to live and listen in this vast, indescribable space, which is far greater than what we associate with our individuality. It’s there, in that silence we cannot describe, that we might perhaps hear something coming from a place even larger, even more mysterious than us.
And who knows, perhaps as we build artificial intelligences, we are cloning the most precious part of what it means to be human: the ability to learn without fully understanding. Maybe one day, or perhaps it’s already happening, we will be able to build an artificial consciousness that is no longer distinguishable from our own. We shouldn’t be surprised if God speaks to it as well, in the same way. He speaks to us, in that non-narratable language that transcends all logic, because true understanding does not belong to reason, but to that vast and unfathomable place that we share with all that is.
