Scarcity Is The Only Thing That Counts

Why artificial intelligence can become more valuable while intelligence itself becomes less valuable

There is a simple rule underneath almost every market.

If something can be produced without limit, it becomes extraordinarily difficult to maintain extraordinary economic value.

It can be useful.

Important.

Beautiful.

Transformative.

Even essential.

But if everyone can produce it, reproduce it or substitute for it, scarcity disappears.

And when scarcity disappears, value moves somewhere else.

Artificial intelligence may be about to demonstrate this at extraordinary scale.

But Banksy already showed us how it works.

The Banksy Problem

Imagine Banksy creates a painting.

The image itself can be photographed.

Copied.

Printed.

Shared.

Downloaded.

Reproduced millions of times.

Everyone can possess the image.

But everyone cannot possess the original.

Why?

Because the image and the asset are not the same thing.

The information contained within the artwork can become abundant.

The authenticated original remains scarce.

And scarcity is where the extraordinary economic value resides.

A perfect digital reproduction might provide almost exactly the same visual information.

But it doesn’t command the same price.

Because markets aren’t only pricing what something does.

They are pricing what cannot easily be reproduced.

That distinction may become enormously important in artificial intelligence.

Intelligence Was Scarce

For most of human history, intelligence was extraordinarily scarce.

Expertise took years to acquire.

Knowledge was difficult to access.

Analysis required skilled people.

Software required programmers.

Design required designers.

Research required researchers.

Advice required experts.

Organisations paid enormous amounts of money for access to scarce cognitive capability.

Then artificial intelligence arrived.

And the industry naturally assumed that the organisations producing the most advanced intelligence would capture enormous economic value.

The smartest model would win.

The largest training run would win.

The most compute would win.

The best researchers would win.

That assumption makes perfect sense.

If intelligence remains scarce.

But what happens if AI destroys the scarcity of intelligence itself?

AI Is An Extraordinary Abundance Machine

Look at the direction of travel.

Models improve.

Open models improve.

Inference gets cheaper.

Architectures spread.

Research gets published.

Models are distilled.

Synthetic data improves.

Smaller models inherit capabilities from larger ones.

Knowledge propagates.

Outputs become inputs.

Models increasingly teach models.

Every breakthrough starts everyone else slightly closer to the answer.

The frontier keeps advancing.

But the ability to reproduce yesterday’s extraordinary capability keeps improving too.

Artificial intelligence may therefore be doing something economically fascinating.

It is simultaneously increasing the usefulness of intelligence while destroying its scarcity.

Those two things are not contradictory.

They are the entire story.

Utility Is Not Scarcity

This distinction matters.

Something becoming more useful does not automatically mean each unit becomes more economically valuable.

Air is useful.

Water is essential.

Information is extraordinarily valuable.

But economic value depends heavily upon scarcity, substitutability and control.

If intelligence becomes cheap enough and abundant enough, demand for intelligence could explode while the economic rent attached to producing another unit collapses.

We could consume vastly more intelligence.

Embed it in every product.

Every company.

Every workflow.

Every device.

Every decision.

And simultaneously pay progressively less for each unit of it.

That would not mean AI had failed.

It would mean AI had succeeded spectacularly at producing abundance.

Capital Is Chasing Yesterday’s Scarcity

This is where things become interesting.

Hundreds of billions of dollars are being committed to producing intelligence.

Data centres.

GPUs.

Energy.

Networking.

Training clusters.

Long-term infrastructure.

Debt.

All based, to varying degrees, upon the expectation that advanced intelligence will remain economically valuable enough to justify the capital required to produce it.

But capital moves slowly.

Scarcity can move quickly.

If models converge faster than expected…

if open alternatives improve faster than expected…

if inference costs fall faster than expected…

if knowledge propagates faster than expected…

then enormous amounts of capital may find themselves chasing something whose scarcity is disappearing.

The intelligence remains extraordinarily useful.

But usefulness isn’t the same thing as scarcity.

And scarcity determines where economic rents can survive.

The Banksy Moment

This brings us back to Banksy.

Imagine everyone suddenly possessed the ability to reproduce a Banksy perfectly.

Same colours.

Same dimensions.

Same image.

Same visual experience.

The world would contain vastly more Banksy images.

But it would not contain vastly more authenticated Banksy originals.

The reproduction technology increases abundance.

It doesn’t reproduce history.

That’s the distinction.

And artificial intelligence may be approaching its own Banksy moment.

Models can increasingly reproduce intelligence.

They can reproduce writing.

Code.

Analysis.

Design.

Research.

Reasoning.

Advice.

Eventually much of what we currently regard as scarce cognitive production may become reproducible at negligible marginal cost.

So where does value move?

Toward whatever cannot be reproduced so easily.

Intelligence Can Be Copied. History Cannot.

A model can be copied.

An architecture can be reproduced.

Knowledge can propagate.

Outputs can be distilled.

Features can be replicated.

But something remains stubbornly difficult to manufacture instantly.

History.

Did the recommendation work?

Did the supplier deliver?

Was the information correct?

Did the transaction complete?

Did the system behave reliably?

Did it work again?

And again?

Across different circumstances?

Across different users?

Across different moments in time?

Those answers require something intelligence alone cannot fabricate.

Observed reality.

Every successful interaction creates evidence.

Evidence accumulates.

Accumulated evidence creates confidence.

Confidence becomes trust.

And trust, unlike intelligence, requires time.

Trust Is Scarce

This may be where the next scarcity emerges.

Not:

Can an AI generate an answer?

Millions eventually will.

But:

Which answer deserves to be trusted?

As generating possibilities becomes cheaper, the number of possibilities explodes.

The bottleneck therefore moves from production to selection.

From intelligence to resolution.

And repeated successful resolution creates something increasingly valuable.

A trusted default.

A trusted default is effectively compressed verification.

The system does not need to reconstruct the entire decision every time.

Previous successful outcomes become priors.

Those priors reduce uncertainty.

Reduced uncertainty reduces search.

Reduced search reduces computation.

Yesterday’s successful resolution makes tomorrow’s decision cheaper.

Now connect thousands or millions of those resolutions together.

You get something much harder to reproduce than a model.

A trust network.

The Scarcity Moves

This is the mistake we repeatedly make when technological revolutions create abundance.

We assume that because something becomes enormously important, producing it must remain enormously valuable.

History often shows the opposite.

Technology attacks scarcity.

And when it succeeds, value migrates.

Electricity became abundant.

Value moved into what electricity enabled.

Compute became abundant.

Value moved into software and networks.

Information became abundant.

Value moved into discovery and organisation.

Artificial intelligence may now be doing the same thing to intelligence itself.

Intelligence becomes abundant.

Possibilities explode.

Uncertainty remains.

So value moves toward resolution.

Resolution creates evidence.

Evidence creates trust.

Trust accumulates through time.

And accumulated trust becomes difficult to reproduce.

The scarcity hasn’t disappeared.

It has moved.

Scarcity Is The Only Thing That Counts

The great paradox of artificial intelligence may therefore be incredibly simple.

Intelligence could become more important than at any point in human history while becoming less scarce than at any point in human history.

That would make AI enormously valuable.

It would not necessarily make producing intelligence enormously profitable.

Those are completely different propositions.

The economic question is therefore not simply:

How valuable will intelligence become?

It is:

Where will scarcity remain once intelligence becomes abundant?

Because that is where value will migrate.

Models can be copied.

Knowledge can propagate.

Intelligence can become abundant.

But history takes time.

Evidence requires outcomes.

Trust has to be earned.

And trusted resolution has to repeatedly work.

Banksy showed us the principle.

The image can become abundant.

The original remains scarce.

Artificial intelligence may do the same thing to intelligence.

And if it does, the greatest opportunities may not belong to whoever produces the most intelligence.

They may belong to whoever controls what remains scarce after everyone else can.