Before the Story

Not with Māori storytelling.

Not with cultural preservation.

Not with some grand plan.

Just a very expensive computer and the obvious question that follows:

What could you actually do with that much machine?

My first thoughts were fairly ordinary. AI video. Generative imagery. Bigger models. Faster processing. All the things you would expect from an absurd amount of compute.

But the more I thought about it, the less interesting the machine became.

Because producing individual AI images or short clips is already relatively easy.

The harder problem is consistency.

A friend of mine had spent a substantial amount making an AI-assisted short film based on a Māori story. Some of the imagery was impressive, but the technology kept doing what generative AI still does remarkably well.

Forgetting what it made five minutes ago.

Faces drift.

Clothing changes.

Buildings subtly reinvent themselves.

The same person can emerge from the other side of a cut looking like a close relative.

And that got me thinking that perhaps the real opportunity is not to begin with the story at all.

Perhaps you build the world first.

Not the world of one particular ancestor or one famous event.

The world people inhabited.

What they wore.

What they built.

How a settlement sat against the land.

The shape and scale of a whare.

The materials.

The tools.

Waka pulled onto the shore.

Cooking.

Firelight.

Weapons.

Vegetation.

Weather.

The way people moved through those spaces.

The thousands of little visual details that tell you where and when you are before anyone has said a word.

That is a very different problem from generating impressive pictures.

It is building a visual language.

And if that language could be built accurately enough, consistently enough and with the right people involved, something interesting happens.

You no longer have to choose one important story and somehow find the enormous budget required to construct an entire historical world around it.

The world is already there.

Or at least the capability to create it is.

Then, when the right person has the right story and the right opportunity presents itself, the barrier to telling it becomes dramatically smaller.

That is the part that interests me.

Because I do not think the technology should decide what stories get told.

I am not particularly interested in feeding famous Māori names into a machine and asking it to manufacture mythology for us.

The stories should come from people.

From whānau.

From hapū.

From iwi.

From storytellers.

From whoever has the knowledge, the connection, the authority or simply the beautiful little story that deserves some light.

The technology should provide the stage.

And perhaps that stage does not always need to hold something monumental.

Someone might eventually use it to tell an important historical account.

Someone else might tell a tiny story about a family.

A child following an adult down to the water.

Someone repairing something.

Preparing food.

A conversation outside a whare.

People laughing.

People arguing.

Nothing historically significant happening at all.

And strangely, I think those ordinary moments might be some of the most interesting.

History tends to leave us the exceptional.

Chiefs.

Battles.

Migrations.

Disasters.

Great speeches.

The things important enough to survive.

What disappears much more easily is everything around them.

What Tuesday looked like.

What the morning felt like.

What sat beside the doorway.

What people did while nothing important was happening.

That world is where the stories lived.

AI may eventually give us a way to reconstruct enough of that world that storytellers are no longer forced to begin every project by inventing it from scratch.

But that only works if the world is believable.

And believable is harder than beautiful.

AI is extraordinarily good at approximation.

Ask it for something vaguely historical and Polynesian and it will confidently give you something cinematic that may also be complete nonsense.

That is precisely the problem.

The challenge would be narrowing the range of things the technology is allowed to get wrong.

Photographs where they exist.

Museum collections.

Archaeology.

Oral histories.

Written accounts.

Landscape.

Language.

Knowledge held by people who understand the difference between something that merely looks Māori and something that belongs in this particular place and time.

Then you begin building.

Not one giant model that supposedly knows everything.

Pieces.

Clothing.

Objects.

Architecture.

Landscapes.

Materials.

Ways of moving through spaces.

Each corrected and refined until it becomes reusable.

In that sense, it starts to resemble a film studio.

Except instead of constructing a village for one production and pulling it apart when filming is finished, the backlot is digital.

Every project adds something.

Every correction improves something.

Every object built properly once becomes available to the next storyteller.

The world becomes richer because people use it.

And that is where my original ninety thousand dollar computer thought begins to look a little silly.

Compute matters, obviously.

There will always be a faster GPU and a more ridiculous workstation to lust after.

But hardware may be the least difficult part of this.

The harder assets are knowledge, access, reference material, trust and people willing to say:

No, that is wrong.

That would not be there.

That does not belong with that.

Try this instead.

Those things cannot simply be solved by ordering another graphics card.

And perhaps that is the real possibility here.

Not that artificial intelligence learns how to tell Māori stories.

I am not sure I particularly want it to.

But that the cost of creating the world around those stories becomes low enough that far more Māori storytellers can.

Build the world first.

Then see who walks into it.