#604 Human Incentives will dictate AI

#604 Waiting on the Robots? Four Layers That Decide What Happens Next

Resistance to AI follows incentives, and incentives can be redesigned.

One of my recent posts was a song about waiting for the robots to take me away. It’s a joke, but the feeling behind it is real. I hear it from neighbours, friends and family: AI is coming, it’s smarter and faster than we are, and there’s nothing to do but wait.

I don’t think that’s true. Here’s how I see it.

My prediction

People will resist the development of superintelligence, except for the versions that help themselves. We already see it. Ask someone about “AI” and they worry. Ask about the assistant they use every day and they love it. The abstract thing is scary, but the personal thing is useful.

That isn’t hypocrisy. It’s how people have always behaved, because the way the world works hasn’t changed. I think four layers explain most of it.

Layer 1: Physics

Everything starts with limits: energy, materials, time. A superintelligence still needs electricity, chips, cooling and land. Those can be measured, and anything measurable can be questioned. Local examples: grid capacity issues, waste water infrastructure, or the new Canada Labs facility on Blair. What's more, laws of physics are absolute, unlike flakey human derived laws.

Layer 2: Incentives

Inside those limits, people respond to rewards and costs. Want to predict what a company, a council or a neighbour will do? Don’t read the headline, ask who gains and who pays. Fear says nobody can predict this technology. Incentives say we can predict a lot of it.

Layer 3: Power

When incentives collide, whoever controls the scarce thing sets the terms: the budget, the agenda, the data, the permit. Local example:developers who actually decide, and who is not in the room.

Layer 4: Institutions

Rules, budgets and procedures outlast the people who wrote them. They lock in whatever balance of power existed at the time. That’s why the unglamorous stuff, like zoning, procurement and public consultation, matters more than most people think.

Why this should make you less afraid

Technology enters at layer 1, but it gets shaped at layers 2, 3 and 4. Those layers are made of people, and people can be organized, persuaded and held to account. The machine isn’t deciding our future. Humans with incentives and power are deciding it, through institutions, in public view if we choose to look.

What you can do

• Measure the limits. Ask for the numbers on power, water and cost.
• Align the rewards. Ask who benefits, who pays, and whether that can change.
• Organize at choke points. Find who decides and show up.
• Use the slow levers. Zoning, budgets and procurement rules are where outcomes get locked in, and they’re open to input early.

In upcoming posts I’ll apply these four layers to decisions facing Ottawa. For now, if you find yourself waiting on the robots, ask which layer you can push on today.

Kayaky

Why the Future of AI Follows the Same Four Layers as Everything Else

A model for being less afraid and more useful.

Most of the fear about superintelligence has the same shape: something enormous is coming, it will do what it does, and the rest of us will watch. That feeling of powerlessness is understandable. It is also, I think, mistaken about where the leverage actually is.

My thesis is simple. People will resist the development of superintelligence, except for the versions that help themselves. That isn’t cynicism. It is how people have behaved with every powerful technology, and it points to something we can use.

The model: four layers

Outcomes in human affairs, including outcomes involving AI, are shaped by four layers. They run from the most fixed to the most changeable.

1. Physics. What is possible: energy, compute, materials, time. Not negotiable. A technology that needs a power plant needs a power plant.
2. Human incentives. What people want and respond to: security, status, money, belonging, advantage. Stable across generations. Incentives can be redirected, but not wished away.
3. Power dynamics. Who holds leverage over whom: owners, regulators, voters, employees, rivals. These shift, but slowly, and rarely because someone asked politely.
4. Institutions. The rules, organizations, and habits that channel the three layers above: laws, procurement rules, courts, standards, oversight bodies. This is the layer people actually design.

What “people don’t change” really means

I don’t mean that society never changes. Institutions, norms, and technologies change constantly. I mean that human motivations are stable, while the structures that channel them are not. People in 1926 and people in 2026 want broadly the same things: safety, advantage for themselves and their families, a say in what happens to them.

That is why the thesis holds. People will welcome AI that helps them and resist AI that threatens them, and each group will fight to shape the rules accordingly. The outcome depends less on what AI “wants” and more on who designs the channels, and for whose benefit.

Why this should make us less fearful

Fear tends to attach to the first two layers. “It will do whatever it wants” is a claim about physics and incentives, and no ordinary person has leverage there. But the model says that outcomes are decided higher up, in the layers where people have always had a say, however imperfectly.

The question isn’t “what will AI do to us?” It is “who writes the rules, and who is in the room?”

What each layer asks of us

• Physics: Learn the constraints. Energy, compute, and supply chains limit what is possible and who can do it. Knowing them separates real risks from science fiction.
• Incentives: Ask who benefits from each version of AI. Follow the incentive and you can usually predict the behavior.
• Power: Notice who has leverage and who doesn’t. Build leverage where it is missing: through coalitions, informed voters, and public pressure.
• Institutions: Show up. Rules about how AI is bought, used, audited, and explained are being written now, mostly in places that look boring.

An example close to home

Consider a city deciding whether to use an AI system to prioritize service requests. The technology (layer one) is the least interesting part. The real questions sit above it. Who benefits if the system works (incentives)? Who controls the data and the vendor relationship (power)? Is there a public procurement process, an audit requirement, a way for residents to challenge a decision (institutions)? The same technology produces a very different city depending on how those answers come out.

The takeaway

People won’t change, so the model asks us to stop hoping they will and start designing for who they are. Fear says the outcome is already decided. The four layers say the decisive layer is the one we can still shape, and the work is local, specific, and mostly unglamorous.

Related: my song “Waiting on the Robots” (post #594).