Contemporary discourse surrounding the enigma of consciousness frequently finds itself polarized, caught between two dominant viewpoints. One perspective, known as computational functionalism, posits that thought processes can be entirely encapsulated by abstract information processing. Proponents of this view contend that any system exhibiting the correct functional organization, irrespective of its underlying material composition, should inherently possess consciousness. Conversely, biological naturalism asserts a diametrically opposed stance, arguing that consciousness is inextricably linked to the unique characteristics of living brains and bodies. This school of thought emphasizes that biology is not merely a passive vessel for cognition but an integral component of cognition itself. While both of these frameworks offer valuable insights, the persistent impasse suggests that a crucial element in understanding consciousness remains elusive.
Introducing a novel perspective, our recent publication advocates for an approach termed "biological computationalism." This designation is deliberately provocative, intended to invigorate and refine the ongoing dialogue. Our central thesis is that the prevailing computational framework is fundamentally flawed, or at the very least, inadequately suited to the intricate reality of how brains actually function. For an extended period, there has been a pervasive inclination to conceptualize the mind as akin to software executing on neural hardware, with the brain "computing" in a manner analogous to conventional digital computers. However, this comparison falters when confronted with the actual architecture of biological brains, which are not structured as von Neumann machines. Forcing this analogy leads to the proliferation of imprecise metaphors and fragile explanatory models. To develop a robust understanding of how brains compute and what would be necessary to engineer minds in alternative substrates, we must first broaden our definition of what "computation" truly entails.
Biological computation, as we delineate it, is characterized by three fundamental attributes.
The first defining characteristic is its hybrid nature, seamlessly integrating discrete events with continuous dynamics. Neurons generate action potentials, synapses release neurotransmitters, and neural networks transition through states that resemble distinct events. Simultaneously, these events occur within a perpetually evolving physical milieu, encompassing fluctuating voltage fields, chemical gradients, ionic diffusion, and time-dependent conductivity changes. The brain operates not as a purely digital system, nor as a simple analog machine, but rather as a multi-layered construct where continuous processes exert influence over discrete events, and in turn, these discrete events reconfigure the continuous background. This intricate interplay unfolds in a continuous feedback loop.
Secondly, biological computation is intrinsically scale-inseparable. In traditional computing paradigms, a clear demarcation often exists between software and hardware, or between a "functional level" and an "implementation level." Within the brain, however, such a clean separation proves untenable. There is no discernible boundary where one can definitively isolate the algorithm from its physical underpinnings. Causal relationships permeate multiple scales concurrently, spanning from the behavior of individual ion channels and dendritic structures to the dynamics of neural circuits and the activity of the entire brain. These levels do not function as independent modules arranged in a hierarchical fashion. In biological systems, alterations to the "implementation" invariably affect the "computation" itself, due to their profound and inextricable entanglement.
The third essential feature of biological computation is its metabolic grounding. The brain operates under stringent energy limitations, and these constraints profoundly influence its structural organization and functional capabilities across all levels. This is far more than a mere engineering consideration; these energy constraints actively shape what the brain can represent, its learning mechanisms, the stability of information patterns, and the coordination and routing of data. From this vantage point, the pervasive and tight coupling across different scales is not a consequence of gratuitous complexity but rather a sophisticated energy optimization strategy that underpins robust and adaptable intelligence within severe metabolic limitations.
When these three core attributes are considered collectively, they lead to a conclusion that may initially seem counterintuitive to those accustomed to classical computing principles. Computation within the brain is not merely an abstract manipulation of symbols, nor is it solely about the movement of representations governed by formal rules, with the physical medium relegated to the status of "mere implementation." In the realm of biological computation, the algorithm and the substrate are one and the same. The physical organization does not simply facilitate the computation; it constitutes the computation itself. Brains do not execute programs in the conventional sense; rather, they are a specific type of physical process that computes through its temporal unfolding.
This re-conceptualization also illuminates a significant limitation in how many contemporary artificial intelligence systems are characterized. Even highly sophisticated AI systems primarily engage in function simulation. They learn mappings between inputs and outputs, often demonstrating impressive generalization capabilities, but the underlying computation remains a digital procedure executed on hardware designed for a fundamentally different mode of operation. In stark contrast, brains perform computation within physical time. Continuous fields, ionic flow, dendritic integration, localized oscillatory coupling, and emergent electromagnetic interactions are not peripheral biological "details" that can be disregarded in the pursuit of an abstract algorithm. In our view, these are the foundational computational primitives of the system, serving as the mechanisms that enable real-time integration, resilience, and adaptive control.
This perspective does not imply that consciousness is exclusively confined to carbon-based life forms. We are not advocating for an "all-or-nothing" biological requirement. Our assertion is more nuanced and pragmatically focused. If consciousness, or cognition exhibiting mind-like qualities, is contingent upon this specific form of computation, then it may necessitate a biological-style computational organization, even if instantiated in novel substrates. The pivotal consideration is not the literal biological nature of the substrate, but rather whether the system embodies the requisite form of hybrid, scale-inseparable, and energetically (or more generally, metabolically) grounded computation.
This reframes the objective for any endeavor aimed at constructing synthetic minds. If brain computation cannot be divorced from its physical realization, then the mere scaling of digital AI may prove insufficient. This is not to diminish the potential for digital systems to achieve greater capabilities, but rather to highlight that capability alone is only a partial solution. The more profound risk lies in potentially optimizing for the wrong criteria by enhancing algorithms while leaving the fundamental computational ontology unchanged. Biological computationalism suggests that the creation of truly mind-like systems may require the development of entirely new classes of physical machines whose computation is not structured as software operating on hardware, but rather distributed across multiple levels, dynamically interconnected, and shaped by the constraints of real-time physics and energy dynamics.
Consequently, if our aim is to achieve something akin to synthetic consciousness, the central question may shift from "What algorithm should we implement?" to "What kind of physical system is necessary for that algorithm to be inseparable from its own intrinsic dynamics?" What specific features are indispensable, including the interplay of hybrid event-field interactions, multi-scale coupling devoid of discrete interfaces, and energetic constraints that actively sculpt inference and learning, such that computation becomes an intrinsic property of the system itself, rather than an abstract description layered upon it?
This represents the fundamental paradigm shift advocated by biological computationalism. It redirects the challenge from the pursuit of the correct program to the identification of the appropriate kind of computing matter.
About the Author
