Can Rocks Compute? What It Means for a Physical System to Be a Computer

A Strange Question

Here’s a weird thing to think about. You’re sitting at a table. In front of you is a computer and a rock. The computer runs programs, plays videos, lets you type words and see them appear on screen. The rock just sits there.

But is the rock really not computing? Or could it be that the rock is doing some kind of computation, just a very boring one that nobody notices? And if the rock isn’t computing, what exactly does the computer have that the rock lacks?

This might sound like a silly question. Obviously a rock isn’t a computer—you can’t play Minecraft on it. But philosophers noticed something troubling. If you look closely enough at anything in the physical world, you can find a way to describe it as performing a computation. A rock gradually heating up in the sun, for example, goes through a sequence of states (cooler, warmer, warmer still) that could be mapped onto the states of a simple computer program that just alternates between 0 and 1. Under the right description, the rock is “computing” just as much as your laptop is.

This is a real problem, not a joke. If everything is a computer, then saying “the mind is a computer” becomes meaningless—it’s like saying “the mind is made of matter.” Of course it is, but that doesn’t tell you anything interesting. So philosophers have been arguing for decades about what genuine physical computation is, and whether rocks (or brains, or anything else) really do it.

The Problem of Implementation

Let’s start with something you already know. There’s a difference between abstract math and physical stuff. A Turing machine (the mathematical model of a computer) is an abstract thing—you can’t drop one on your foot. It’s defined by rules, states, and symbols on a tape that never runs out. But your laptop is a physical thing made of silicon, wires, and electricity. Somehow, the physical laptop implements the abstract computations. The laptop’s physical states (voltages in circuits) correspond to the Turing machine’s abstract states (0s and 1s), and as the voltages change, the abstract computation happens.

This is called the problem of computational implementation: how does a physical, concrete system manage to be the same thing as an abstract, mathematical computation? It’s not obvious that this should even be possible.

The simplest answer is also the most troubling one. Here’s philosopher Hilary Putnam’s proposal, later called the Simple Mapping Account: a physical system implements a computation if you can find a way to map the physical states of the system onto the computational states defined by the computation, in a way that the physical state transitions match the computational state transitions. That’s it. Just draw a map from physical stuff to computational stuff, and if the physical stuff changes in a way that follows the computational rules, the system is computing.

And here’s the problem: you can do that for anything. Remember the rock heating in the sun? You can map its temperature states onto the 0 and 1 states of a simple two-state calculator, and as the rock warms up, its temperature changes match the calculator’s state changes. By Putnam’s definition, the rock is a perfectly good computer.

This conclusion is called pancomputationalism—the view that literally every physical system is a computer. Most philosophers think this can’t be right. If everything computes, then nothing specially computes. The word loses its meaning. But if pancomputationalism is wrong, then we need a stricter definition of what counts as genuine computation. Let’s see what philosophers have come up with.

Stricter Accounts: Making Computation Harder

The Counterfactual Account

One problem with the rock-and-sun example is that the rock’s temperature changes aren’t caused by the rock itself. The rock isn’t producing heat; the sun is heating it. If you put the rock in the shade, it would cool down instead of getting warmer. The rock doesn’t have a built-in disposition to go from one temperature to the next the way a computer has a built-in disposition to go from one computational state to another.

The Counterfactual Account says that to genuinely compute, a physical system must support counterfactuals: “if the system were in state A, it would go to state B.” This is different from just happening to be in state A and then state B. The rock fails this test because it’s not the rock’s nature to go from temperature T to T+1; that depends entirely on whether the sun is shining.

This account rules out the rock. But it still lets in a lot of other things. The motion of planets, for instance, supports counterfactuals: if the Earth were in a slightly different position, its trajectory would change accordingly. Does that mean the solar system is a computer? Some philosophers say yes, but they call this limited pancomputationalism—everything computes, but different things compute different computations, and the computations they perform are objective facts about them.

The Causal and Mechanistic Accounts

The Causal Account goes further: genuine computation requires that the physical states cause each other in the right way. A computer’s state transitions are caused by the internal workings of the machine. The rock’s temperature transitions are caused by an external energy source. Most philosophers think this gets closer to the right answer.

The Mechanistic Account is even more specific. Developed by philosopher Gualtiero Piccinini, it says that a computing system is a special kind of functional mechanism—a system whose parts are organized to perform specific functions. A digestive system is a functional mechanism (its parts work together to digest food), and a computer is a functional mechanism too (its parts work together to process information). But the computer’s function is to manipulate medium-flexible vehicles—basically, physical states that can carry different kinds of information depending on how they’re organized.

Here’s the key insight: a genuine computer has parts that are supposed to do certain things. If a computer malfunctions and gives the wrong answer, it’s still a computer—it just failed to perform its function correctly. But a rock can’t malfunction, because it doesn’t have a function at all. The rock isn’t supposed to do anything. This is why, according to the mechanistic account, rocks don’t compute: they aren’t the right kind of mechanism.

The Semantic Account

Here’s another idea. When you use a computer, the symbols it manipulates mean something. The bit pattern “01100001” in your computer stands for the letter ‘a’. According to the Semantic Account, there is “no computation without representation”—computation just is the processing of meaningful symbols.

This account is popular in philosophy of mind because it seems to distinguish brains (which obviously manipulate meaningful thoughts) from rocks (which don’t). But it raises a tough question: what makes something a representation? A rock’s temperature doesn’t represent anything—it’s just a temperature. But the voltage levels in your computer do represent numbers and letters. How do we tell the difference? Philosophers disagree violently about this. Some say representation is an objective property of physical systems; others say it’s just something we project onto them. If it’s the latter, then the semantic account collapses back into the simple mapping account, and we’re right back where we started.

The Universe as a Computer

Some physicists take pancomputationalism even further. They argue that the universe itself is fundamentally a computer—not just something that can be described by a computer, but something that is a computer at its most basic level.

The simplest version says the universe is a giant cellular automaton: a grid of cells, each of which can be in one of a few states, updating its state in discrete steps based on its neighbors. This is the view of physicist Edward Fredkin and mathematician Stephen Wolfram, among others. The idea is that if you zoom in far enough, space and time aren’t continuous—they’re made of tiny discrete steps, like pixels on a screen. And the rules for how one step follows the next are exactly the rules of a computation.

A more sophisticated version says the universe is a quantum computer—not a classical one that manipulates 0s and 1s, but one that manipulates quantum states (qubits) that can be in superpositions. This version has some support from physics, since quantum mechanics does look a lot like a quantum computation in some interpretations.

But both versions face a serious objection. Physical stuff has properties that purely computational stuff doesn’t have. A computation is just a pattern of abstract states. But a rock isn’t just a pattern of abstract states—it’s concrete matter that takes up space, has mass, and can hit you in the head. Saying the universe is “just” a computation seems to leave out the hard question: where does the concrete physical stuff come from? This is the same problem the ancient philosopher Pythagoras faced when he said “all is number.” It’s a beautiful idea, but it’s very hard to make it work.

The Physical Church-Turing Thesis

There’s another way this question matters. The Church-Turing Thesis (named after logicians Alonzo Church and Alan Turing) says that anything that’s intuitively computable can be computed by a Turing machine. Most mathematicians accept this. But what about physical computation? Could there be a physical system that computes things no Turing machine can compute? This would be a hypercomputer.

Several proposals for hypercomputers have been made. One involves a machine that performs each step in half the time as the previous step, so it can do infinitely many steps in a finite amount of time. Another involves a black hole that allows an observer to see the result of a computation that took forever to complete. A third uses quantum mechanics in a clever way to solve problems that are unsolvable by classical computers.

So far, none of these proposals has produced a working physical system. The relativistic hypercomputer requires a specific kind of spacetime (Malament-Hogarth spacetime) that we don’t know if our universe contains. The quantum hypercomputer requires infinite precision in setting up and measuring the system, which seems physically impossible. The infinitely accelerating machine requires parts that shrink forever, which violates atomic physics.

This is why most philosophers accept what’s called the Modest Physical Church-Turing Thesis: any function that a physically constructible, reliable, usable computing system can compute is also computable by a Turing machine. In other words, even if there are weird physical processes that could compute uncomputable functions, we can’t actually use them as computers. For all practical purposes, physical computation stays within the bounds that Turing and Church discovered back in the 1930s.

Back to the Rock

So: does the rock compute? The answer depends on which account you accept. Under the simple mapping account, yes—everything computes. Under the counterfactual account, maybe not, but lots of other things still do. Under the causal or mechanistic account, probably not—the rock doesn’t have the right kind of causal structure or function. Under the semantic account, definitely not—rocks don’t manipulate meaningful symbols.

The interesting thing is that nobody has settled this debate. Philosophers still argue about which account is correct. And part of what makes it hard is that the question isn’t just about rocks and computers—it’s also about brains. If the brain turns out to be a computing system (and many cognitive scientists think it is), then which account of computation we accept determines what kind of computing system the brain is, and what it can and can’t do. If the brain is a computer in the same sense that your laptop is a computer, that tells us one thing. If it’s a computer only in the sense that a rock heating in the sun is a computer, that tells us something very different.

The rock question, as silly as it sounds, turns out to be a question about what we are.

Appendix

Key Terms

Term	What it does in this debate
Pancomputationalism	The view that every physical system performs computations, which threatens to make the claim “this system computes” trivial
Simple Mapping Account	The earliest and most permissive definition of physical computation: any system that can be mapped onto a computational description counts as computing
Counterfactual Account	A stricter definition requiring that a system’s state transitions support “what would happen if” statements, not just describe what did happen
Mechanistic Account	The view that genuine computing systems are functional mechanisms with parts organized to manipulate medium-flexible vehicles
Functional Mechanism	A system whose components are organized to perform specific functions, like a digestive system or a digital computer
Medium-flexible Vehicles	Physical states that can carry different kinds of information depending on their organization, allowing the same computation to run on different hardware
Hypercomputer	A physical system that could compute functions no Turing machine can compute; none has been successfully built
Church-Turing Thesis	The claim that anything intuitively computable can be computed by a Turing machine; the physical version extends this to physically constructible computers

Key People

Alan Turing — British mathematician who in 1936 invented the Turing machine and proved that some mathematical problems cannot be solved by any algorithm; he also helped break Nazi codes in World War II.
Hilary Putnam — American philosopher who proposed the simple mapping account of computation and later argued that it leads to pancomputationalism, which he thought was a problem for the theory of mind.
Gualtiero Piccinini — Contemporary philosopher who developed the mechanistic account of computation, arguing that computing systems are functional mechanisms that manipulate medium-flexible vehicles.
John Searle — American philosopher who argued that whether something implements a computation depends on how an observer interprets it, which supports pancomputationalism (though he was actually trying to argue against strong AI).

Things to Think About

If rocks don’t compute, at what point does a collection of stuff become a computer? Is a single transistor a computer? A handful of sand? A brain? Where’s the line, and who draws it?
The mechanistic account says computers have functions—they’re supposed to do certain things. But what makes something a function? Is a rock that happens to be useful for holding paper performing the function of a paperweight? Is that the same kind of function a computer has?
Suppose someone builds a hypercomputer that works. Would that change what we think the human mind is capable of, or would it just tell us something about physics? If the brain turns out to be a hypercomputer, how would we know?
If the universe is a computer, what is it computing? And who or what is running the program? Could there be a computer that isn’t running for anyone or anything?

Where This Shows Up

Artificial intelligence: The question “does this AI really think?” depends partly on whether the AI’s computations are the same kind of thing as human mental processes, which is exactly what these accounts try to decide.
Limits of science: The hypercomputation debate asks whether there are physical limits to what humans (or our machines) can ever know or calculate.
Everyday technology: When your calculator gives a wrong answer, is it still computing? Mechanistic accounts say yes—it’s miscomputing. Simple mapping accounts say it’s just computing something else.
Digital physics: Some physicists seriously investigate whether the universe is fundamentally a computer, which would mean that physics is a branch of computer science. This sounds like science fiction, but it’s an active research area.