The Water Is in the Machine Too

I set out to create an image.

That’s how this started.

A simple companion to a blog about stress—about how we live in it like fish in water, rarely questioning its source, its cost, or its quiet shaping of our lives.

I thought the image would be the easy part.

I was wrong.

What I was trying to create was simple, at least in concept:

A world underwater.

People moving through it as if nothing was different.

No panic. No awareness. Just life.

Because that’s what stress feels like.

Not an event.

An environment.

But as we began generating the images, something else emerged.

Something I wasn’t expecting.

Or maybe something I was—but hadn’t yet seen this clearly.

The First Distortion

The people were wrong.

Not in obvious ways.

In subtle ones.

The kind you can feel before you can name.

There were too many faces that didn't reflect the abundance of diversity in this world.

Too many bodies that reflected a narrow standard of beauty.

Too many compositions that felt… curated.

Even when I asked for diversity, what came back was a version of it.

Balanced. Aesthetic. Palatable.

Not true.

And so I adjusted.

I asked for more representation.

More bodies. More cultures. More reality.

And the system responded.

But not cleanly.

Overcorrection and Erasure

In some images, diversity became performative.

In others, it disappeared again.

Black bodies were underrepresented.

Indigenous presence was almost invisible.

Same-sex relationships were either missing or exaggerated.

People were grouped into neat clusters—couples, families, units—as if human life must always be organized into something recognizable.

As if randomness, aloneness, or quiet co-existence had no place.

And then I saw it.

Not just in the images.

But in the process itself.

The Bias Beneath the Output

The system wasn’t failing.

It was revealing.

Revealing the data it was trained on.

Revealing the patterns it had learned.

Revealing the defaults it returns to when not carefully guided.

A world where:

• Whiteness is still the baseline

• Thinness is still the norm

• Youth is still preferred

• Heteronormativity is assumed

• Connection is staged

• Diversity is curated, not lived

In other words: the water is in the machine too.

And Here’s the Part That Matters Most

This wasn’t just a technical exercise.

It was emotional.

Frustrating.

At times, disorienting.

Because I could see the gap between what is true and what was being generated—and closing that gap required constant correction.

Constant noticing.

Constant refusal to accept “close enough.”

And that’s when it hit me.

This is what it feels like to live inside systems we didn’t design.

Systems that shape reality in ways that are almost invisible—until you try to see differently.

The Image We Finally Created

We got there, eventually.

An image where:

• People moved through the world without knowing the water was there

• There was no central figure

• No forced connection

• No curated diversity

• Just humanity, dispersed, present, and real

It wasn’t perfect.

But it was honest.

Why This Matters

Because AI is not separate from us.

It is trained on us.

On our images. Our histories. Our preferences. Our biases.

It doesn’t invent distortion.

It reflects it.

And if we’re not paying attention, it will quietly reinforce the very patterns many of us are trying to unlearn.

So What Do We Do?

We don’t reject the tools.

We engage with them consciously.

We question what we see.

We notice what’s missing.

We ask better questions.

We refuse easy answers.

Because the goal is not perfection.

It’s awareness.

A Final Reflection

I started this trying to illustrate stress.

But what emerged was something deeper.

A reminder that:

We are not only shaped by the systems we live in

We are now shaping the systems that will shape us

And if we want a different world—

We have to be willing to see the one we’re in.

Even when it’s uncomfortable.

Even when it’s subtle.

Even when it looks… almost right.

Because almost right is how the water stays invisible.