It's Time for AI to Deliver the Goods

When we talk about AI adoption, most people mean efficiency. Making existing processes faster. Swapping one process stack for another.

Here's the thing. With what has been borrowed against AI, efficiency does not come close to settling the account.

Because that is what this is. A debt. We have directed much of the world's capital into a single technology and away from everything else it could have funded. That capital was not free. It was borrowed against a promise.

And capital is only part of what was drawn down. The data centres behind AI run on staggering amounts of energy and water. Their electricity demand is on track to double by 2030 — to roughly 945 terawatt-hours a year, more than all of Japan consumes. Power that could light homes and run hospitals. Fresh water, by the millions of litres, poured through cooling systems to keep the servers cold. Water that could grow food and fill reservoirs. Every model trained, every query answered, draws down resources that are already scarce — and not available to people who need them.

So the account is large, and it is still open.

Against it stands the promise. One day, we were told, AI would help cure cancer, reduce hunger, clean up pollution, make life more viable for everyone. That was the note we signed. That was the collateral we put up.

So where is the payment going? Not toward those problems. Toward compute centres, power generation, GPUs, memory. The machinery of something its builders call AGI.

Here is why that was never going to end any other way.

Cybernetics has a law for it: the law of inevitable ethical inadequacy. Stated plainly — if you do not require a system to be ethical, you will not get one. You get a system that optimizes for the goals you specified, at the expense of the ones you did not. The ethical. The human. The good.

And those goals are set by us. Not by the machine — by the people who build it and fund it. The system does what it is told. We are the ones doing the telling. So the inadequacy is not the machine's failing. It is ours.

Look at what we made a requirement. Capability. Scale. AGI. Then look at what we left as a promise. Cancer. Hunger. Pollution. A better life for everyone.

A promise is not a requirement. And a system built to hit the goals we set will spend every borrowed dollar, every watt, every litre on hitting them harder — not on the good we never required of it.

So the default was built in from the start. In both senses of the word. The outcome defaults to what we specified, and the debt defaults unpaid — because repayment was never written into the terms.

What is being built looks like a future without us. Not one where we do better. That is not a betrayal of the goal. It is the goal, working exactly as specified.

The bill does not disappear because we spent it elsewhere. If AI is going to cost us this much — our capital, our power, our water, the things we need to live and grow our society — then the good cannot stay a promise. It has to become a requirement. Written into the terms. Measured in what we were told AI would deliver. Cancer. Hunger. Not process cycle times.

That is what is owed. Specify the good, or it will not come. That is the only way AI ever delivers the goods.