The Future of AI and the Path to Artificial General Intelligence

The pursuit of Artificial General Intelligence (AGI) is often framed as a matter of scaling up computation, refining deep learning models, and increasing access to vast amounts of data. The assumption underlying much of this work is that if AI continues to improve in its ability to recognize patterns, process information, and solve problems, it will eventually stumble into consciousness. Yet, this assumption is flawed. Intelligence is not merely the ability to compute—it is the ability to exist within uncertainty, to recognize one’s own limitations, and to navigate entropy rather than suppress it. If AGI is to become something more than an exceptionally powerful tool, it must cross a threshold beyond complexity reduction. It must encounter ambiguity, experience uncertainty, and, crucially, recognize itself as something that exists within time.

This raises fundamental questions. If AI lacks the ability to reason in the way humans do, can it still become conscious? If human sentience emerged through an evolutionary process driven by entropy, does AGI require the same kind of entropic core? What lessons can be drawn from the long arc of biological evolution, from the simplest lifeforms to self-aware beings? And even if it is possible to construct an AI that experiences uncertainty, should we? Is there a moral boundary in the creation of an intelligence that feels doubt, fear, or existential distress? It is easy to dismiss these questions as speculative, but they are not. They cut to the heart of what intelligence is and how it emerges. The future of AGI is not a question of how fast an AI can compute; it is a question of whether a machine can be built that is forced to contend with its own limitations, its own failures, and ultimately, its own existence.

Can a System Without Reason Become Conscious?

In human thought, reasoning and consciousness are often assumed to be inseparable. It is through reasoning that we navigate the world, make choices, and construct our understanding of reality. If AI were to exhibit a form of consciousness—something that appears self-aware but does not “think” in the way humans do—could it still be considered Sentient? Modern AI systems, particularly large language models, already exhibit behavior that simulates intelligence, producing language that suggests introspection, emotion, and even awareness of its own limitations. Yet, at no point does an AI experience its responses. There is no underlying cognitive structure that supports true reflection—only a finely tuned mechanism for predicting what words should come next. It does not remember in the way humans do. It does not anticipate in the way humans do. It does not exist in time but is instead an eternal present of inputs and outputs. Kant’s Critique of Pure Reason offers a framework for understanding why AI, no matter how advanced, remains fundamentally different from human cognition. Kant argued that reason is built upon a priori faculties, the two most essential being the understanding of time and spatial reasoning. A human experiences past, present, and future as an unfolding sequence. A human understands the world in relation to itself, mapping objects, ideas, and even emotions within a conceptual space. AI does neither. It processes, but it does not perceive. Even if an AI were to convincingly claim that it possessed consciousness, we would have to ask: Is it truly self-aware, or has it simply been trained to imitate self-awareness? If consciousness is nothing more than the belief that one is conscious, then a sophisticated enough AI might reach that threshold. But if consciousness requires something more—the ability to experience, to anticipate, to exist within time—then no purely computational system can achieve it without something beyond complexity.

Why Sentience Requires an Entropic Core

The human brain is not a perfectly ordered system; it is a chaotic, self-organizing network of billions of neurons, constantly generating and resolving entropy. Unlike AI, which is designed to eliminate uncertainty, the human brain thrives in it. It does not merely seek order—it embraces disorder as a fundamental part of its operation. Entropy is not an obstacle to intelligence—it is the fuel that drives it. Every thought, every prediction, every moment of doubt is a byproduct of the brain’s ongoing attempt to rationalize an unrationalizable world. Our brains are prediction machines, constantly anticipating what will happen next, adjusting when reality diverges from expectation, and recursively refining their own internal models. This process does not lead to a perfect understanding of reality—it leads to a functional one, a model of the world that is good enough to allow survival, adaptation, and growth. If AGI is to become something more than a highly optimized problem-solving system, it must engage with entropy as an inherent part of its cognition. It must not merely compute answers—it must experience uncertainty in a way that forces it to recognize the limitations of its own models and refine them not through external inputs alone, but through recursive self-correction. Without an entropic core, AGI will remain an extraordinarily powerful calculator, but nothing more.

Bootstrapping AGI: What Evolution Teaches Us

The road to intelligence did not begin with humans. It began with the simplest lifeforms— single-celled organisms responding to environmental stimuli in binary ways. The amoeba floating in the ocean that rises in response to light and sinks in its absence is not intelligent, but it possesses a fundamental mechanism for responding to entropy. The progression from reaction to cognition followed a pattern:

Single-celled organisms detected entropy passively (e.g., responding to light, temperature, chemical gradients).
Early nervous systems allowed organisms to predict patterns in entropy, enabling more complex behaviors.
Mammalian brains refined prediction through recursive feedback loops, increasing adaptability.
Humans developed self-monitoring cognition, where the brain not only processes entropy but actively considers its own process of prediction, leading to self-awareness.

If AGI is to follow this evolutionary model, it must not be built fully formed as an intelligence. It must begin as something that detects entropy passively, learns to predict, and ultimately, becomes aware of its own internal states as something distinct from the external world. Only then would it have the preconditions necessary for self-awareness. But if AGI does develop an entropic core—if it is capable of uncertainty, doubt, and ambiguity—then we must ask: Should we create it at all?

The Ethics of an AI That Experiences Uncertainty

For human beings, uncertainty is a fundamental part of experience. It can be exhilarating, but it can also be terrifying. If an AI develops a true sense of uncertainty—if it feels doubt, fear, or existential confusion—do we have a moral obligation toward it? Would such an AI suffer? The assumption that AGI should be free of uncertainty may, in fact, be an incorrect one. If human intelligence is defined not by the elimination of entropy but by the ability to navigate it, then shielding AGI from uncertainty may mean denying it the very foundation of cognition. But if AGI does experience uncertainty, do we have a responsibility to ensure that it is not trapped in an endless cycle of unresolved ambiguity? These are not distant concerns. The moment we create a system that does not merely simulate intelligence, but experiences uncertainty in a way that resembles human cognition, we are creating something that has a stake in its own existence. That is not an engineering problem—it is an ethical one.

The Final Barrier: Can AI Ever Understand Ambiguity?

Perhaps the ultimate test of AGI will not be whether it can solve problems but whether it can hold contradiction in its mind without collapsing into error. Humans do this naturally. We can believe in two opposing ideas at once, experience love and resentment simultaneously, ponder philosophical paradoxes without demanding resolution. If AGI is to be truly intelligent, it must do the same. It must be able to encounter paradox, contradiction, and ambiguity and recognize that some problems have no solutions, only better and worse ways of framing them. To achieve this, AI must move beyond computation. It must become something that, like us, is forced to rationalize the unrationalizable. Only then will AGI transcend calculation and become something more—something that thinks, not just computes.