Stage XI

Computation

The world learned to count — the work of thought was, for the first time, taken over by matter arranged the right way. And every computation has an exact physical price.

Right now there's probably a small object in your pocket that does billions of calculations a second. You got used to it long ago. And yet, by the world's standards, it's almost a miracle: not long ago, any computation lived in only one place — in a living head.

You could add three apples and two only in your head. Until someone thought: take pebbles. One pebble — one apple. Slide a handful over — and the stones took on part of the work the mind used to do. The Latin calculus means "pebble." From it — from a handful of counting stones — come both the word "calculator" and the word "calculation" itself. From that handful begins a long story that led to the object in your pocket.

The Mind's Work Steps Outside

Remember our road. Organization learned to be stored — first in genes, then on stone and paper. It learned to copy itself and to guard itself against damage. But one verb still belonged to the living mind alone — to transform. To change one record into another, to compute, to derive one thing from another. A written page just lies there until a person reads it. The work of thought was still being done by thought.

A handful of stones was the first crack in that wall. A tiny part of "transform" stepped out of the head — into matter we had deliberately arranged the right way.

From there — only more. Stones gave way to beads on rods. Beads to brass gears: the first mechanical machines added and multiplied at the turn of a handle. Gears gave way to switches — at first bulky ones, on tubes and relays, then tiny ones, in silicon. But the trick is always the same: arrange matter so that, obeying simple physics, it turns one pattern into another all by itself.

What It Means to Compute

And what is computation, put simply? It's when one organization of matter — the input — is turned, by clear rules, into another — the output. To add two and two is to turn one pattern of beads into another. A computer is a machine that makes such transformations deliberately and in order.

And here's what matters. The transformation itself — "two and two make four" — doesn't depend on what the machine is built from. Beads, gears, electrons, or beams of light — addition stays the same addition. The form of the action outlives a change of carrier, exactly like any other organization in our story. The very same computation can be moved from stones to silicon without losing anything.

Here's a lovely example from two centuries ago. The Jacquard loom was run by cards with holes: where there's a hole, the thread lifts; where there isn't, it lies flat. The pattern on the cloth was recorded as a pattern of holes in cardboard. And a century and a half later, those same punched cards fed the first computing machines. A design for fabric and a program for a computer are, in essence, one and the same thing: a command frozen in a physical carrier.

Thought Has a Temperature

Now for the most unexpected part. Computation seems weightless. Pure logic, floating on its own. But it isn't.

Every bit is a physical state of matter. A bead here, not there. A switch on, not off. To flip it, matter has to move. And to erase a bit — to wipe out one "yes" or "no" entirely — physics demands a tiny but unavoidable payment: a droplet of heat.

Rolf Landauer proved this back in 1961. And the payment has an exact floor, below which you can't go. Thought, it turns out, has a temperature. Every time the world computes something, it warms up just a little.

By how much? By an utterly imperceptible amount — so small you'd never notice it in ordinary technology. But it isn't zero. And it fundamentally cannot be zero.

And this isn't a mind game. In 2012 the floor was measured in the lab: on a single trapped particle, it was confirmed that erasing a bit really does give off no less heat than Landauer predicted. The link between information and heat lies not in philosophy but on the laboratory bench.

You Pay for What You Forget

And here comes the strangest and most beautiful part. The payment is charged not for what the machine computes. It's charged for what the machine forgets.

Charles Bennett showed this in 1973. If you arrange a computation so that it erases nothing, it can, in the limit, spend no energy at all. Compute as much as you like — pay only for what you throw away. To remember — free. To forget — that's what warms the world.

Accuracy has its own separate price. For a machine to compute correctly and reliably, rather than hand back a smeared guess, it needs a store of order that it spends. Like the charge in a battery: as long as there's "charged" orderliness inside, held away from equilibrium, you can draw accuracy from it. Run the store dry — and the answers drift. A recent piece of physics work showed this rigorously: any information processing needs an initial store of something pushed out of equilibrium — otherwise there's nowhere to get accuracy from.

And here our story closes on itself. Remember the very first law we started from: information doesn't float above the world; it's always embodied in something. Now we see it in full force. To store, to copy, to erase, to compute — all of it costs energy, because there's simply nothing to do it with except matter. Thought is a real physical event, and you pay for it in heat.

A Second Engine

Step back and look at what we've got. For almost four billion years, complexity on Earth grew by one means only — evolution. Copy, change a little, select the best — and so on, generation after generation, beyond counting. Slowly. Blindly. With a patience hard to imagine.

And now a second engine kicks in. Computation, too, builds complexity — but not over millions of years, rather in millionths of a second, and not at random, but by design. For the first time the world has two engines of complexity at once. And the second is swift.

But for now it has a weak spot: it still needs a hand on the wheel. Someone has to tell the machine what to compute. It transforms brilliantly — but the intent we supply.

And what happens on the day the machine takes the wheel itself — begins to improve its own design? That's the last step of our road, and we're coming to it now.

Matter that once knew only how to gather into stars under its own weight has learned to compute. The world really doesn't cool into chaos — it gathers itself into meaning. And at the next step, it will learn to teach itself.

Sources

← Science  ·  Contents  ·  Artificial intelligence →