Is the brain a computer? A debate that has lasted for 80 years

I. Summary of the Core Content

This article delves into a niche yet highly significant academic controversy at the very foundation of the AI era: Roman Bleute, a leading French neuroscientist, has written a book that is virtually unknown to the general public. The book, titled *The Theoretical Brain*, ranks over 150,000 on Amazon’s overall bestseller list but has received prominent recommendations from top academic journals such as *Nature*. It directly challenges the “fundamental consensus” that has dominated the AI industry and the field of neuroscience over the past 80 years—that the brain is a biological computer. Starting from the underlying logic of Darwin’s theory of evolution, Bleute exposes the flaws in four mainstream assumptions: that the brain performs calculations, has internal representations of the external world, specializes in processing information, and operates through prediction. He argues that equating the human brain with a man-made computer essentially imposes an engineering logic of “a designer creating a product according to a blueprint” on a biological entity that evolved entirely through natural selection, which can lead the research direction in these fields astray. This debate, which has been ongoing for nearly 80 years, may seem to be about scientists refining definitions, but it actually has a profound impact on how we will develop AI in the future, how we define intelligence, and even how we distinguish between humans and machines. It represents one of the most critical underlying issues of the AI era.

---

II. Popular Explanation of the Key Points

1. Why has this obscure book, which sells so few copies, caused such a stir in the professional community?

This is not just a casual argument from a layperson. The author, Bleute, is the director of the Paris Institute of Intelligent Systems and Robotics and is recognized as a top neuroscientist in the field; even *Nature* has dedicated a book review to it. Its impact is akin to someone publicly pointing out that the foundation of a building might be incorrectly designed. For decades, AI research, brain-computer interfaces, and brain-inspired chips have been based on the assumption that the brain’s workings are identical to those of a computer. It’s like everyone building a 100-story building on the premise that the foundation is solid, only for a seasoned engineer to emerge and say that we have fundamentally overlooked the essential difference between living organisms and man-made machines. The value of this book is not to overthrow all existing research, but to bring the overused and rigid metaphor of “brain = computer” back under scrutiny, giving theoretical research, which is often overshadowed by a flood of experimental data, a renewed focus. Currently, neuroscientists are busy collecting data for papers and few take the time to question the fundamental logic of this metaphor.

2. Bleute’s critique of the mainstream consensus in simple terms

His arguments are easy to understand for laypeople:

• First, if the brain is programmable, then who writes the programs? The “weights” in AI (which can be thought of as the strength of connections between neurons—strong connections allow signals to be transmitted quickly, while weak ones do not respond no matter how loudly they are signaled) can be set by engineers at will. However, evolution has no predetermined goals and does not write instructions; only individuals that cannot survive are eliminated. Claiming that evolution writes programs introduces the idea of a divine designer, which contradicts Darwin’s theory of evolution.
• Second, the human brain does not have an “external user.” The parameters of a typical computer are set at the factory and do not change during operation, but the connections between neurons in the brain are constantly changing based on what you see and do. There is no “user” to adjust these parameters; you can’t call something a computer if it has no controls or user.
• Third, even single-celled organisms like paramecium, which lack even a single neuron, can avoid dirty substances and move towards food sources, and they can learn and change their behavior. This undermines the premise that intelligence must be derived from neural network calculations.

3. The four camps in the 80-year debate about “is the brain a computer” are now clearly defined

The debate began in 1943 with the creation of the first mathematical model of a neuron and has lasted for 80 years. The current positions are clear:

• The “Computational Camp,” which constitutes over 90% of research resources, believes that the brain is a biological computer. This camp is responsible for developments in brain-computer interfaces, whole-brain simulations, and brain-inspired AI, and all current AI achievements are based on this perspective.
• The “Improvement Camp” does not oppose the idea that the brain is a computer but argues that the previous understanding was incorrect. They believe the brain actively predicts what will happen next and adjusts itself accordingly, making it more efficient than passive processing.
• The “Opposition Camp,” led by Bleute, argues that the brain is a living, evolving entity, not a man-made one. Intelligence arises from the interaction between the brain and the environment. While they have not yet proposed a testable or engineering-based alternative, they have not provided a convincing alternative, limiting their influence.
• The “Indifferent Camp” of AI engineers believes that the debate is a waste of time since there is no consensus on the definition of a computer. By a broad definition, anything that can perform calculations is a computer, so the brain fits this category. By a narrow definition, a computer is man-made, with a user and adjustable parameters, and the brain does not meet these criteria. They focus on the functionality of the models, not whether the brain is a computer.

4. This seemingly unrelated academic debate actually determines the future of AI

Bleute’s opposition is not about the brain’s computational capabilities but about the assumption that the brain is like the old-fashioned computers of the 1950s, with separate hardware and software that can be precisely modified by an external user. Interestingly, the most advanced AI models are moving away from the definition of traditional computers and towards the characteristics described by Bleute. No one can fully explain why a model with billions of parameters produces certain outputs, and engineers cannot control all its behaviors individually. These models develop capabilities that were not anticipated by their creators. The relationship between humans and AI models is no longer one of a programmer giving instructions and a machine executing them; it has become a symbiotic system that evolves together. This debate suggests that our previous definition of “human” may be flawed. Humans are not pre-programmed biological machines, and the boundary between humans and machines is not black and white. If an AI with a physical body, a life experience, and deep connections to human society emerges, it may be considered intelligent. There may never be a definitive answer to this question, but precisely this exploration makes the AI era so fascinating.