The Complexity of Prediction

I have always liked the idea of predictability.

As a child, I remember playing with dice and trying to understand why I could not get a six every time.

I would try to repeat the movement.

Same direction.
Same strength.
Same motion.

But the result changed.

At the time, I knew nothing about probability.

I was simply trying to understand what decided the result.

Was the answer hidden somewhere in the movement of the dice?

How could I predict the result?

Later, I learned about probability.

I liked the idea.

But I also realized that not knowing the next result did not mean knowing nothing.

With the dice, each side has a probability.

That still did not tell me what the next result would be.

But now I knew something.

Throw the dice once, and anything could happen.

Ten throws could still look irregular.

But after many throws, a pattern began to appear.

Each number moved closer to its expected frequency.

The individual result remained uncertain.

But the overall behavior became somehow stable.

Then things became more interesting with two dice.

Now the possible sums were not equally likely.

Seven, for example, could be created in more ways than two or twelve.

Repeat the experiment many times and the results begin to form a pattern.

The values concentrate around the middle.

Seven appears most often.

Six and eight appear slightly less.

Then five and nine.

And so on.

Randomness was not the absence of structure.

I could see that there were patterns inside uncertainty.

That changed my question.

What could I actually predict?

Or perhaps:

Which rules did I need to discover to predict the result?

Later, I learned physics.

And it gave me a very different view of the world.

Objects did not simply move.

They followed rules.

For me, it started with the story of Newton and the apple.

A falling object accelerates because of gravity.

Its path depends on its position, its velocity and the forces acting on it.

And the same principles could describe the movement of planets.

The more I learned, the more powerful this idea became.

In some cases, this worked extraordinarily well.

It suggested a very different possibility.

Perhaps the future was not random at all.

Perhaps it was calculable.

But nature does not move from one fixed state to another.

It changes continuously.

A falling object changes position.

Its velocity changes.

And the rate at which its velocity changes is acceleration.

The system is constantly evolving.

This difference between something discrete and something continuous became important to me.

The dice gave me separate outcomes.

One result.

Then another.

But motion was different.

It was happening continuously.

Every moment was connected to the one before it.

This is where calculus and differential equations became important.

At first, the ideas seemed abstract.

But eventually I understood what those equations were really doing.

They were describing change itself.

Instead of simply saying where a system is, a differential equation describes how its current state is changing.

A small step forward in time.

Then another.

Then another.

What looks like continuous motion can be described as an evolving system.

For something simple, like a falling object, this can work beautifully.

But then another question appears.

What happens when billions or trillions of things interact at the same time?

The equations had not disappeared.

The physical laws were still there.

But the systems became more complicated.

One object alone behaves in one way.

Add another object, and now they interact.

Each one affects the other.

And the result is different from considering each one alone.

Add another.

And another.

The number of interactions grows.

One change affects another part of the system.

That part changes something else.

And those changes can eventually return and influence the original motion again.

Everything begins influencing everything else.

The problem is no longer simply finding the rule.

The problem becomes following what those rules produce when the system has many possible interactions.

This changed the way I understood prediction.

Knowing the laws governing a system does not automatically make its future easy to calculate.

And that opened an entirely new world for me.

Air.

Air looks almost empty.

It is not.

A gas contains an enormous number of molecules moving continuously.

They travel.

They collide.

They change direction.

They exchange energy.

Something as ordinary as the air inside a room contains a microscopic world in constant motion.

Much more than I could imagine.

And properties that seem simple at our scale have a different meaning underneath.

At this scale, predicting the exact trajectory of every individual molecule becomes impractical.

There are simply too many of them.

So the way I thought about the problem had to change.

Instead of asking where every molecule was, I needed to begin thinking about the behavior of the system as a whole.

The individual particles were still there.

But from all those individual movements, new collective properties emerged.

The same idea applies to water and other fluids.

Instead of tracking each molecule independently, fluid dynamics describes the material as something continuous.

Now we can imagine a value at every point in space.

At one location, the fluid has a certain velocity.

A certain pressure.

A certain temperature.

A certain density.

Move somewhere else, and those values are different.

Together, they form fields.

And these fields evolve through time.

Fluid dynamics tries to describe how.

There are equations that describe parts of this process.

They connect motion, pressure, viscosity and forces.

The mathematics can sometimes be written compactly.

But the behavior it produces can be anything but simple.

Watch smoke rise.

At first, the motion can look smooth.

The stream moves upward in an almost organized way.

Then something changes.

The flow bends.

A vortex appears.

Then another.

Large structures break into smaller ones.

Those smaller structures interact again.

Soon, the smooth motion has become turbulence.

What is striking is that the physical rules have not changed.

The fluid is still obeying the same underlying laws.

But the interactions inside the flow create extremely complex behavior.

Small differences can propagate through the system.

This is where chaos enters the story.

A small variation in velocity, temperature or pressure changes what happens nearby.

That change affects another region.

Which affects another.

Over time, two systems that began almost the same can evolve into very different states.

The system can still be deterministic.

But that does not mean it remains easy to predict far into the future.

Then the problem becomes much larger.

The atmosphere is a fluid.

The same ideas now operate across an entire planet.

Air moves because of differences in pressure and temperature.

Water evaporates.

Clouds form.

Energy arrives from the Sun.

Oceans exchange heat with the atmosphere.

The Earth rotates.

Mountains redirect flows.

Many different processes interact continuously.

Weather forecasting becomes an enormous problem in fluid dynamics.

We measure the current atmosphere as accurately as possible.

Then numerical models calculate how those quantities should evolve.

But our measurements are never perfect.

And the equations are extremely complex.

Small uncertainties in the current state can grow as the forecast moves further into the future.

So many simulations can be run with slightly different starting conditions.

At first, they may remain close.

Later, they begin to spread.

And that spread itself contains information.

It tells us something about the range of possible futures.

And about how confident we can be in the prediction.

To me, prediction seemed to mean one thing:

Knowing what would happen next.

Over time, that definition became less relevant.

Probability showed me that uncertainty could still contain structure.

Physics showed me that nature follows rules.

Mathematics showed me how systems evolve continuously.

Complex systems showed me that knowing the rules does not guarantee a perfect forecast.

Fluid dynamics made that complexity visible.

And eventually I learned that prediction can become too complicated to describe as one certain future.

The question changes.

Instead of asking only:

What will happen?

I begin asking:

What can happen?
Which outcomes are more likely?
How confident are we?
How does that confidence change with time?

Prediction becomes less about discovering one certain future.

It becomes a way of understanding the structure of possible futures.