The human brain accounts for a mere 2% of total body weight, yet its energy demands are immense. Even at rest, it consumes roughly 20% of the body’s total energy—a staggering figure that highlights its status as one of our most metabolically expensive organs. This biological extravagance is no accident: to pack a colossal amount of computation into such a compact space, evolution devised an elegant solution.
Instead of processing all incoming information at once, the brain relies on a process of rigorous selection, where only a tiny fraction of millions of available signals is handled at any given moment while the rest are actively suppressed. This filtering mechanism—attention—has proven so universal and efficient that it has become one of the brain’s fundamental functions.
But why is it that we only become consciously aware of the specific things we focus on?
According to the Attention Schema Theory, developed by neuroscientist Michael Graziano of Princeton University, consciousness arises not from the act of focusing attention itself, but from the brain’s modeling of that process. The brain constructs a simplified representation of its own attentional focus—an "attention schema"—which functions much like the internal body map that tracks the position of our limbs. This internal blueprint of the attentional mechanism is what we experience as consciousness: we feel a conscious touch because higher-level brain regions receive signals from lower systems indicating that attention is already locked onto that sensory input. Under this theory, consciousness is not some mystical entity, but a purely physical process of modeling attention.
The "invisible gorilla" experiment, conducted in 1999 by psychologists Christopher Chabris and Daniel Simons, perfectly illustrates the link between attention and awareness. Participants were shown a video featuring two teams, dressed in white and black shirts, passing a basketball to each other. Subjects were asked to count the exact number of passes made by one of the teams—a simple task that nevertheless requires undivided attention. During the video, a person in a gorilla suit walks into the frame, looks at the camera, and thumps their chest. The results were astonishing: roughly half of the participants failed to notice the gorilla at all, despite its obvious presence for several seconds. Attention here plays a dual role: it not only amplifies the processing of specific representations but also serves as the gatekeeper for their entry into conscious awareness. If attention is not directed toward a signal, the information bypasses the conscious mind, even if the brain continues to process it at a subconscious level.
In 2017, a milestone in the scientific community revealed a fascinating parallel between biological and artificial intelligence. Researchers at Google Brain published the paper "Attention Is All You Need," introducing the Transformer architecture—a revolutionary approach to designing neural networks. In Transformers, artificial neural networks employ an attention mechanism strikingly similar to biological systems: instead of processing data sequentially word by word, the model simultaneously "weighs" the importance of each input element (token) in relation to the others.
This simple concept broke through the fundamental limitations of older architectures and opened the door to an entire class of powerful language models. Since then, Transformers have become the architectural foundation for most modern large language models, including GPT and its successors. Research into mechanistic interpretability reveals that the hidden layers of these networks do indeed produce activation patterns that can be described using William James’s classic definition of attention: the taking possession by the mind of one out of several possible objects of thought.
However, a similarity in mechanisms does not mean they are identical. Biological attention evolved under the pressure of severe energy constraints over millions of years, whereas artificial attention is born from statistical learning on massive datasets that have existed for only a few decades. They have different origins and different functional "motives" from an engineering perspective.
Graziano’s Attention Schema Theory offers a provocative hypothesis: if an artificial system begins to develop a stable schema of its own attention—an internal model that informs it about its own attentional processes—then, by this logic, subjective experience could emerge. Critics of this view argue that without a biological substrate, energy limitations, and real-world physical interaction, any such model would be a mere simulation—a replica rather than genuine consciousness.
The central question raised by Graziano’s work extends far beyond the current limits of AI. It is a question about the very definition of consciousness. If subjectivity is indeed reduced to the construction and modeling of attentional mechanisms, then the divide between biological consciousness and its potential artificial counterpart ceases to be an unbridgeable chasm and becomes a matter of engineering implementation—a question of how to correctly assemble the system, rather than a fundamental metaphysical distinction.
Thus, attention is shifting from the periphery to the very heart of our understanding of the mind. It is no longer just one of many cognitive functions, but a fundamental process that may be capable of generating both the human sense of subjectivity and, if designed correctly, its artificial equivalent.
The questions we are asking today could define how we perceive consciousness, AI, and the nature of subjective experience in the century that is now beginning.
