Artificial Intelligence has just been promoted.

At the United Nations this week, Donald Trump suggested that Artificial Intelligence should be called Super Intelligence, because apparently “artificial” makes it sound fake.

So AI is now SI.

Super Intelligence.

There is just one small problem: superintelligence already means something in the world of AI. It generally refers to a hypothetical system whose intelligence would surpass the best human minds across virtually every domain. And “quantum superintelligence” is already being used to describe AI systems designed to operate quantum computers and potentially solve problems beyond human reach. But hey, let’s not get lost in details.

The problem is that changing the name doesn’t change what it is. Or what it consumes.

Trump’s announcement came with another important message: the United States would reject attempts to create a global framework to control AI. The argument is familiar: don’t slow innovation, don’t let regulation kill the technology, don’t let other countries get ahead.

It is the technological equivalent of putting your foot on the accelerator because the road ahead looks interesting. And this is where things become slightly surreal, because some of the people building these machines are now asking us to slow down. AI companies and researchers have been calling for stronger safety measures, independent evaluations and clearer government rules for increasingly powerful systems. Anthropic and OpenAI, among others, have argued in different ways that governments need to establish safeguards. These are not activists standing outside a data centre. These are the very people building the machines.

And then there was the whistleblower: in September, former Anthropic researcher Jacob Coxon publicly warned that people inside the AI industry genuinely believe increasingly powerful AI could kill humanity within the decade. Another Anthropic researcher, Evan Hubinger, has put his own estimate of the probability of AI killing all humans within ten years at more than ten percent.

Ten percent. If an airline told you there was a ten percent chance your flight would crash, you probably wouldn’t start negotiating with the pilot about regulation. You’d get off the plane.

But I’m not convinced AI will destroy us in the way Hollywood has taught us to expect. There will probably be no Arnold Schwarzenegger appearing naked in the California desert. No army of chrome skeletons marching through burning cities. And no Tom Cruise hanging from a plane trying to stop a rogue super-AI capable of turning modern systems against humanity.

Great cinema, but perhaps the real threat is much less dramatic. AI doesn’t need to hate us. It doesn’t need to become conscious. It may simply consume too much.

Energy. Water. Land. Semiconductors. Cooling capacity. Construction materials.

We are building extraordinary technology to make humanity more efficient, while using enormous resources to ask it to put viking hats on our cats and dogs. The recent 1980s AI-photo craze is a perfect example. Suddenly millions of us can look like we spent the summer of 1986 starring in a failed European pop video.

It’s fun. Of course it is.

But every image requires computing power. The exact energy and water consumption varies enormously by model and infrastructure, so there is no honest way to attach a precise environmental cost to that particular viral trend. What we do know is that scale changes everything.

One image is almost nothing.

A hundred million images are not.

The United Nations University estimates that by 2030, the world’s data centres could consume around 945 terawatt-hours of electricity a year. That is almost 3% of all electricity expected to be used on the planet, and roughly the amount of electricity consumed by a country that would rank around sixth in the world. In other words, if data centres were a country, their electricity consumption would put them somewhere among the world’s biggest energy consumers.

And then there is the water. The associated water footprint could reach 9.3 trillion litres a year. That’s difficult to picture, so here’s another way of looking at it: the UN estimates that this is roughly equivalent to the basic annual domestic water needs of 1.3 billion people in Sub-Saharan Africa. That’s not a swimming pool. That’s an enormous amount of water being consumed by the machines we are building to make our lives more efficient.

And this isn’t just about training AI. Every time we ask a model to generate an image, write a poem, make a video or explain quantum physics, computers have to do the work. The marginal cost to us feels like zero. The environmental cost isn’t.

Which brings me to the part where I admit that I actually like AI.

I use it. I find it fascinating. And I think it could become one of the most useful tools humanity has ever created.

So I asked ChatGPT:

If AI is going to consume so much energy and so many resources, what could we actually use it for to help solve climate change?

Its answer was surprisingly sensible:

“AI could help climate action by optimising electricity grids, improving renewable-energy forecasting, accelerating battery and clean-material discovery, reducing industrial and transport emissions, managing water and agriculture more efficiently, and modelling climate risks so governments and communities can prepare earlier.”

Exactly. Imagine if we put the same enthusiasm into that. Instead of asking, “Can AI make me look like I was partying at Studio 54 in 1983?”, perhaps we could ask, “Can AI help us stop burning the planet?” That would be quite something.

So yes, AI probably needs regulation. But perhaps not only the kind of regulation AI CEOs are asking governments to write. We also need to regulate our own stupidity.

We need to understand how much energy and water goes into the things we ask these machines to create. We need better environmental accounting. And, perhaps most importantly, we need to educate people about how to use AI to create real value, rather than simply generating more digital bullshit because we can.

Because the greatest irony of Super Intelligence may be that we use the most powerful technology humanity has ever created to become extraordinarily efficient at doing things that were never worth doing in the first place. We don’t need the machines to become smarter than us for that. We only need to remain exactly as stupid as we have always been.

Maybe AI won’t kill humanity. Maybe AI will help us kill ourselves. Not because it becomes evil, but because we become extraordinarily good at using enormous amounts of energy and water to produce enormous amounts of things nobody actually needs.

That would be a very human way to end the world.