The Danger of Thought

The question of whether computers can think is not simply a technical question. It is also a neuropsychological, philosophical, linguistic, and social question. A computer can process information. It can detect patterns. It can generate language. It can solve problems. It can simulate reasoning. It can even appear to explain itself. But none of these abilities, by themselves, proves that a computer thinks in the way a human being thinks.

From a neuropsychological point of view, the central distinction is this: the human brain does not merely process information; it lives inside a body, regulates that body, experiences needs, forms intentions, attaches meaning to perception, remembers emotionally, acts in the world, suffers consequences, and continuously updates its understanding through embodied experience. A computer network, including today’s artificial intelligence systems, operates on representations, probabilities, weights, tokens, training data, and optimisation functions. It may produce outputs that resemble thought, but resemblance is not identity.

This distinction matters because modern AI is becoming extraordinarily good at imitating the external signs of thinking. It writes in coherent sentences. It uses causal language. It says “I think,” “I reasoned,” “I considered,” and “my conclusion is.” It can break a problem into steps, compare alternatives, and produce confident recommendations. To the human mind, these are powerful social signals. We are biologically inclined to attribute mind to anything that speaks in a human-like way. But the fact that a system can produce the language of thought does not mean that it possesses thought.

Alan Turing understood part of this problem as early as 1950. Rather than asking directly whether machines can think, he proposed what became known as the Turing Test: if a machine’s responses could not be reliably distinguished from a human’s in conversation, perhaps the practical distinction would become less important. The Stanford Encyclopedia of Philosophy notes that Turing regarded the question “Can machines think?” as too ambiguous to be useful in its ordinary form. The test therefore shifted attention from inner experience to observable performance.

That shift was brilliant, but it also created a lasting confusion. Passing as intelligent is not the same as being conscious. Producing intelligent behaviour is not necessarily the same as possessing understanding. A mirror can reflect a face without having a face. A map can represent a city without living in it. A language model can produce the external structure of thought without necessarily possessing the internal conditions from which human thought arises.

John Searle’s famous Chinese Room argument made this point in philosophical form. In the thought experiment, a person who does not understand Chinese sits in a room and follows a rulebook for manipulating Chinese symbols. To outsiders, the responses appear meaningful. But inside the room, there is no understanding of Chinese — only symbol manipulation. Searle’s argument was designed to challenge the claim that running the right program is sufficient for genuine understanding.

Whether one accepts Searle’s conclusion fully or not, the argument remains highly relevant to modern AI. Large language models do not begin with lived meaning. They begin with patterns in data. They are trained on vast collections of text, images, code, and other digital material. They learn statistical relationships between elements of that material. When prompted, they generate outputs by predicting what is likely to come next within a given context. The result can be astonishingly fluent, but fluency is not the same as comprehension.

The human brain is different in almost every relevant respect. It is not an isolated symbol processor. It is a living, biological control system embedded in a body. It receives information from the external world through the senses, but it also receives continuous information from the internal world of the organism: hunger, fatigue, pain, breath, heartbeat, hormonal state, immune activity, bodily tension, and emotional arousal. Modern neuroscience increasingly emphasises the role of interoception — the brain’s monitoring of internal bodily states — in emotion, decision-making, selfhood, and conscious experience. Reviews of brain–body physiology describe behaviour as tightly synchronised with bodily needs and physiological regulation, not merely with abstract information processing.

This matters because human thought is not detached calculation. It is need-guided, value-laden, emotionally weighted, memory-dependent, socially shaped, and action-oriented. When a human thinks, the brain is not simply asking, “What is the next likely word?” It is asking how to survive, how to thrive, how to connect, and how to act. Thought is the instrument of life, not merely the product of processing.