Organism versus machine.

“Your brain runs on 12 watts. AI needs billions of watts to do the same thing.”

Every few weeks a claim resurfaces that goes something like this.

It sounds interesting but it is also totally wrong.

The adult human brain uses closer to 20 watts, accounting for around 20 percent of the body’s resting energy use, despite being only about 2 percent of body mass. That figure is well established in neuroscience.

But the real problem is not whether the number is 12 watts or 20. It is the idea that AI “needs billions of watts”. It doesn’t.

No AI system burns gigawatts to read a sentence or generate a response. That kind of power output is what you expect from a nuclear power station. The number usually comes from muddling together data centre capacity, peak infrastructure limits, and long-run training costs, then pretending they describe a single cognitive act (I will come back to why so many speakers, authors and linkediners do this)

They do not.

Subscribe

AI energy use varies wildly depending on hardware, scale, and task. Training large models can draw megawatts over time. Running them uses far less. None of this maps cleanly onto how our brains work.

And that is the deeper issue. Brains use metabolic energy to sustain a living organism (remember ATP from your GCSEs?) AI uses electrical energy to perform computation. Comparing watts across those systems does not tell us anything meaningful about intelligence.

The brain reads this text, attaches meaning, filters it through memory, emotion, bias, and experience, and keeps you alive while doing it. Heart rate, breathing, hormones, posture, attention. All in parallel; which is pretty amazing.

We do not understand how that all works; not even close.

What we do have is AI that is very good at pattern recognition and matching at scale. This is definitely impressive engineering. But not in any way natural intelligence.

So framing this as an “efficiency race” misses the point entirely.

This is not bicycle versus a freight train. Rather it’s organism versus machine.

We are not closing in on natural intelligence any time soon and we are still trying to understand what it is. Judging by how confidently we repeat bad numbers and worse metaphors, we might want to focus on that first.

What do you think we are really comparing when we compare brains to machines?

#scottspeaks #futureliteracy #messymiddle