Why the Theory That the Brain Is Like a Computer Is Wrong

Think of your computer for a moment. According to one theory about the human mind, the brain is like a computer and the mind is the software that runs it.

This theory is called computationalism or, sometimes, computer functionalism. The theory has been widely accepted as a sort of orthodoxy in academia for quite awhile.

It seems plausible on account of the advances in artificial intelligence. In many realms, computation can exceed human mental capabilities. Artificial intelligence software runs on physical computers so we find it easy to assume that the mind is somehow physically running on the brain in the same way.

However, this intuition has a fundamental flaw. It breaks down when we analyze it logically.

What’s the flaw?

Just because we can replicate human thought with a computer, does not mean that the brain is capable of the same kind of processing as a computer. In fact, the success of artificial intelligence could give evidence against the mind being a product of the brain.

John Von Neumann (1903–1957), who invented the architecture used by most computers, wrote a book, The Computer and the Brain (Yale University Press 1958). He compared the computer and the brain by their electrical and computational properties. He found that the properties we need for very precise, and rapid calculation in computers are not to be found in the brain.

Bringing this back to artificial intelligence, von Neumann’s insight shows why the very success of artificial intelligence undermines the case for computationalism. Artificial intelligence requires enormous amounts of extremely deterministic, precise, rapid, and synchronous calculation. Going to these extremes is the only way we have found that brings computers to the level of the human mind. On the other hand, the brain’s electrical activity is extremely noisy, imprecise, slow, and disjointed. Researchers have tried for decades to use the native characteristics of the brain to create effective artificial intelligence algorithms, and all have failed.

Neural networks vs. neurons

Neural networks are the prime example of this division. The artificial neuron started out as a high fidelity model of biological whale neurons, known as the McCulloch-Pitts neuron.

Found the formula?

The authors used their artificial neuron to model logical formulas, and claimed they’d found the basis of rational thought in the brain’s neural structure. However, these neural networks floundered for decades because there was no effective way to train them except by the intelligent design of engineers.

It is only the recent invention of the backpropagation algorithm — which has no corollary in the brain — that neural networks have become the workhorse of today’s AI revolution. And even so, multiple other non-biologically based innovations have been necessary to bring neural networks architecture to a point that can replicate human artifacts.

So, the pattern we see in artificial intelligence progress is that the movement towards emulating the human mind is also a movement away from copying the brain. That is why artificial intelligence, instead of being evidence for computationalism, is actually evidence against it.

But there is an even more remarkable fact we should realize: The success of artificial intelligence shows that to whatever extent the mind operates like a computer, such a mind cannot be running on the hardware of the brain. At best, the brain is only a receiver for the mind’s computations.