Substrate Objections Resurrected. John Searle's Biological Naturalism in the LLM Era

John Searle’s Chinese Room argument is arguably the most famous philosophical challenge to artificial intelligence. Proposed in his 1980 paper “Minds, Brains, and Programs” in Behavioral and Brain Sciences (Searle, 1980), the thought experiment argued that syntax is not sufficient for semantics. A system can manipulate symbols perfectly according to formal rules without understanding the meaning of those symbols. For decades, the argument stood as a theoretical boundary. The rapid advancement of large language models has forced a re-evaluation of Searle’s position, specifically his broader commitment to biological naturalism.

The Syntax-Semantics Gap in Transformers

Large language models are the ultimate realization of the Chinese Room. They take input tokens, apply complex but purely formal mathematical weights, and produce output tokens. They have no grounding in the physical world and no sensory access to the objects their tokens describe. They are vast mathematical engines operating entirely on syntax.

Despite this, the sophistication of their outputs regularly causes users to attribute genuine understanding. Searle’s argument anticipates this illusion. The person inside the Chinese Room passes the Turing Test by successfully following the rulebook, yet understands nothing of the language. The transformer architecture follows the same principle at an unimaginable scale. It maps statistical relationships between symbols. It never bridges the gap between the symbol and the referent.

Modern functionalists counter that if a system tracks the causal relationships between concepts accurately enough to reason counterfactually and solve novel problems, the distinction between syntax and semantics breaks down. They argue that understanding is just highly structured syntactic correlation. Searle rejected this reduction. He maintained that consciousness and intentionality are biological phenomena, tied to the specific causal powers of the brain.

Biological Naturalism and Causal Powers

Searle’s biological naturalism asserts that mental states are higher-level features of the brain, caused by lower-level neurobiological processes. Just as digestion is a biological process that cannot be achieved by running a computer simulation of an intestine, consciousness cannot be achieved by running a computer simulation of cognition. The physical substrate matters because it possesses specific causal powers that silicon microchips lack.

This framework aligns with the modern biological computationalism arguments documented extensively on this site. Researchers operating in this tradition point to the scale-inseparable nature of biological systems, metabolic integration, and hybrid analog-digital processing as the specific causal powers Searle originally left unspecified. In Google DeepMind’s recent theoretical work, Alexander Lerchner echoed this exact premise, arguing that symbolic computation structurally abstracts away the physical grounding required for experience.

If Searle is correct, current AI development trajectories are pursuing behavioral mimicry rather than genuine mental states. Scaling laws will produce increasingly convincing simulacra. They will never cross the threshold into sentience because they are using the wrong physical material. This raises what researchers call the zombie gap in biological computationalism, leaving developers unable to functionally distinguish between a philosophical zombie and a conscious biological entity.

Evaluating the Substrate Requirement

The state of the AI consciousness debate has fractured over this exact premise. The indicators framework developed by Patrick Butlin and colleagues attempts to evaluate AI systems purely on their functional architecture, implicitly rejecting Searle’s biological exclusivity. They assume that if a system implements the right computational structures, it will possess the associated mental states regardless of the substrate.

To date, the boundary between what requires genuine understanding and what can be statistically approximated remains blurry. The Chinese Room is no longer just a thought experiment. It is the dominant technological paradigm of the decade. Searle’s assertion that syntax never produces semantics remains the primary theoretical hurdle for anyone claiming that conversational AI has achieved true comprehension.