Why AI Creative Control Is Completely Backwards

I asked Leonardo.ai to depict a 15th Century Flemish street view with crowds and a market. What came back stopped me cold.

The architecture was period-accurate, or at least it seemed to be. (Maybe there is a scholar of 15th Century Flemish architecture reading this who could correct me.) The costuming seemed historically consistent. Even the light looked like a Dutch Old Masters painting. It was contrived but convincing.

I had underestimated something fundamental about this tool. Simple concepts get fleshed out with contextual detail that was only implied, creating a textured tapestry of responses that genuinely surprises.

That surprise taught me something many creatives are getting wrong about AI partnership.

The Compounding Effect

The real power emerged when I started layering prompts into the tool. Details compound into more than the sum of its parts.

The creative process becomes enhanced through progressive accretion of detail and adjustment. The dialogue that creates this condition leads to subtle and authored results.

But there is a point where the label of ‘tool’ breaks down. An AI tool is nothing like a paintbrush.

It's more like clutching a robotic arm that has its own choreography and protocols of movement. You can only input moves and manoeuvres: you gesture and imply. The control is less direct. You lead rather than control.

The Dance Metaphor

Your creative vision must allow for the semi-autonomy of the tool. If you want direct control, take up a paintbrush.

The essence of AI collaboration is that detail takes you somewhere unexpected, even while the framework conforms to your vision.

This requires learning to dance with a partner who has their own moves. It implies fluidity and conscious abandonment of rigid control.

Most creatives resist this because they think losing control means losing their authorship and their competitive edge. In fact, they're protecting something that was never real.

Control Was Always An Illusion

Control is an illusion in the practice of any art form. The true artist follows technique and method with exactitude and rigour: this is tight control of process. But this isn't the same as exerting tight control over outcomes.

Art born of rigid control is lifeless and wooden. The dance, the performance, is a letting go.

Research supports this counterintuitive reality. Human-AI teams didn't surpass AI systems operating alone, contradicting popular assumptions about seamless collaboration.

Design and discovery in this process are two sides of the same coin. The Copyright Office concluded that prompts alone don't provide sufficient human control to make users the authors of AI output, but the question remains: does it matter?

The Ethical Elephant in the Room

Right now, I'm preoccupied with the dance while ethics take a back seat. I am yet to reconcile this conundrum satisfactorily.

The training data remains ethically problematic. If you refuse to dance with AI, you avoid sullying your hands with that compromised dimension. But that avoids rather than addresses the ethics question.

Method and technique are separate questions from ethics. Perhaps we need ethically sourced AI tools, like we have ethically sourced food? HarperCollins recently struck deals with Microsoft at $5,000 per title, establishing concrete market value for training data. That seems like a move in the right direction, but the puzzle won't resolve quickly.

Learning The Dance Moves

My advice to creatives still thinking in binary terms of tool versus threat: engage with the technology and see what the dance moves are. See where it takes you.

We need to explore these tools to make informed judgments about them.

The creative professionals who understand AI's semi-autonomous nature will gain advantages over those clinging to outdated narratives about control.

The dance requires conscious abandonment of rigid control. But it's still your dance.