The People Who Switch Off Their 20-Watt Brains

Opening

Dear reader, back when digital cameras first appeared, there was a saying: the best-value camera in the world is the human eye. It was true then. It’s still true now.

These days, I think we can add one more line to that: the best-value intelligence in the world, better than any AI model, is still the human brain. The kind of performance differs, but on efficiency alone, there’s no contest. The brain runs on roughly 20 watts — about as much as a dim incandescent bulb.

So here’s a question: if we all already have such cheap intelligence, why are we building data centers that draw billions of watts? Let me give you the conclusion up front. Machines didn’t win by being smarter than the brain. They won by doing the one thing the brain absolutely cannot: replicate itself. And the real problem starts right after that.

---

The Eye Never Lost to the Camera

People say digital cameras beat the eye, but the story changes once you ask exactly what they beat. Image quality? No. The human eye, with about 6 million cone cells, takes in roughly 1.6 billion bits per second. Most camera sensors can’t hold a candle to that. Even on a price-to-performance basis, the eye keeps winning.

There was exactly one thing the camera won: a scene the eye has seen can’t be pulled out of the head. It can’t be copied, transmitted, or shared identically with someone else. A photo taken with a camera, on the other hand, detaches itself from the lens. It copies perfectly and opens identically on any device. What the eye lost wasn’t a contest of image quality — it was a contest of replicability.

What’s interesting is that even that last wall is starting to crack. Research on scanning the brain and having AI reconstruct the image a person is looking at is getting more sophisticated every year. It still requires a scanner the size of a room, and it’s far better at reading what’s actually in front of the eyes than what’s imagined in the mind. Still, the principle has already been proven.

---

The Brain ‘Loses’ for the Exact Same Reason

Now let’s take this structure up a level. The exact same thing is happening between the brain and AI.

Let’s start with a fact-check. It’s true that the brain runs on 20 watts. But let me clear up one misconception: this isn’t electricity the brain generates — it’s metabolic energy from burning glucose, which is to say, heat. And more precisely, it’s a rough estimate somewhere between 14 and 20 watts, not a spec you could write into a purchase order. No matter how hard you think, consumption rises by at most about 8%. Still, the basic fact that it’s roughly one light bulb’s worth doesn’t change.

So why do we spend billions of watts when such cheap intelligence already exists? Because what AI sells isn’t “smarter than the brain.” What AI sells is a uniform mind that can be copied 1 billion times identically, is always on standby, and can even be audited through logs. The brain can’t do that. It takes 20 years to raise a single one, and that single one can’t be copied.

This is where the cost that 20 watts was secretly hiding comes into view: the energy poured into 20 years of raising and educating that one brain. Here’s why it matters — once a machine is trained, that cost gets amortized across a billion copies, but a brain has to be raised from scratch, one at a time. That’s exactly where the dream of biocomputing — “let’s just grow brain servers and use them” — runs into a wall.

In fact, this field has already reached the commercial stage. Australia’s Cortical Labs became famous for research¹ that trained roughly 800,000 living neurons to play “Pong,” and in 2025 it released CL1, a commercial biocomputer priced at $35,000. Switzerland’s FinalSpark remotely rents out 16 human brain organoids² for $500 a month. All of this is real — either peer-reviewed or actually for sale.

But the scale is embarrassingly small compared to the hype. A single organoid lives for barely 6 months or so, and it can’t be mass-produced uniformly. The phrase “a million times more efficient than silicon” gets thrown around, but that’s the “in principle” possibility the company itself is describing. Once you factor in the incubation equipment and nutrient supply needed to keep it alive, plus a lifespan that needs replacing every 6 months, most of that efficiency disappears. The inability to replicate trips things up here too.

---

But Here’s the Real Twist

If you stop here, the conclusion looks bright. The brain is cheap, and machines only won on replicability, so we can just divide the labor by what each is good at: logic and computation to silicon, massive optimization search to neuromorphic³ chips or quantum computers, and judgment, ethics, and intuition to people. I think this division of labor is, for now, the most realistic answer.

But this division of labor rests on one quiet assumption: that people keep holding onto their own share of the work — judgment. And lately, I’m watching that very assumption start to wobble.

The real danger isn’t machines beating the brain. It’s that because machines cheaply hand out uniform thought — thinking that can be copied anytime — people voluntarily surrender the one asset that can’t be replicated: their own judgment. The cheapest 20-watt brain in the world, and its own owner is the one who stops switching it on.

Picture a concrete scene. Handing a report draft to AI is fine — polishing sentences and organizing material is close to repetition. But if you ship the AI’s conclusion as-is without asking whether it actually fits your situation, what you handed over wasn’t repetition — it was judgment. Code is the same. The moment you paste in AI-generated code without understanding it, what’s left is a system that, later on, nobody can explain why it runs the way it does.

What’s frightening is that this barely shows. Nothing happens on the day you hand off your judgment. If anything, it feels fast and convenient. The bill comes due much later, when the entire organization slides into a state where nobody knows why a decision was made. Individual thinking is a muscle too — leave it unused, and it quietly atrophies.

An essay that sparked this issue paints the scene like this: a future where an external brain runs the city’s power and water while people themselves commute back and forth remembering nothing at all. The author draws a line and calls this science fiction. But to me, it looks like nothing more than a slightly exaggerated version of a present that has already begun.

---

Oswald’s Lens

Honestly, I already made this argument once, back in September 2025. I wrote a book called Homo Brainless: People Who Outsource Their Thinking, subtitled “A warning for modern people who have given up thinking in the age of AI.” Finished in 2024 and released in 2025, the book was, commercially, a complete flop.

search.shopping.naver.comsearch.shopping.naver.com

Back then, almost nobody read it. AI hype was at its peak, so a warning like “don’t hand your thinking to machines” felt deflating. Everyone was only looking at the upside of delegation. What sold well at the time were stories about how amazing AI is, what else it could do, how you should use it. But lately, this book has started getting read again. It’s not because my diagnosis suddenly became right. It’s that the market is only now ready to ask that question.

There’s a pattern I’ve seen over and over while building go-to-market strategies: the better the diagnosis, the more time lag it eats up waiting for the market to be ready. Forecasts that technology will change something are usually right. It’s the timing and the path that are almost always wrong. “Outsourcing thought” was the same. The direction was correct — it just took exactly 2 years for people to actually feel the cost.

So what I want to say isn’t “stay away from technology.” It’s the exact opposite. Already holding the cheapest intelligence in the world in your hands and not switching it on — that’s the most wasteful choice of all.

---

Closing

Let me sum up. The human brain is still the best-value intelligence in the world. What machines beat it on wasn’t performance — it was the single fact of replicability. So the real question isn’t “how smart are machines getting,” but “am I keeping my own irreplicable judgment switched on?”

When you hand something off to AI, I’d suggest making just one distinction: is what you’re handing off “repetition” or “judgment”? Hand off repetition all you like — that’s what machines are good at. But if you’re handing off judgment out of habit too, you’re effectively switching off the cheapest computer in the world.

Have you had a moment recently, while delegating work to AI, where you thought, “wait, this was something I should have judged myself, and I just handed it over”? Tell me in the comments where you crossed that line. In the next issue, let’s map out together, using your examples, where the boundary between repetition and judgment actually sits.

---

💬 If you’ve had a moment where you handed even “judgment” over to AI, tell me in the comments. I’ll fold it into the next issue. 📨 If you know a colleague who seems to be outsourcing their thinking a little too easily these days, quietly pass this along.

📬 If you’d like to receive analysis like this every week, subscribe here I write a weekly newsletter that cross-analyzes technology, economics, and the humanities.

---

References & Further Reading

Primary sources

• Zheng, Jieyu & Meister, Markus, “The unbearable slowness of being: Why do we live at 10 bits/s?”, Neuron, 2025. Link ··· The paper that forms the backbone of today’s issue — the senses take in a billion bits per second, but consciousness processes only 10.
• Kagan, Brett J. et al., “In vitro neurons learn and exhibit sentience when embodied in a simulated game-world”, Neuron, 2022. Link ··· This is the study where neurons learned to play “Pong.” Note that the term “sentience” has drawn pushback in academic circles, so read with that caveat.
• “This $35,000 Computer Is Powered by Trapped Human Brain Cells”, Gizmodo, 2025. Link ··· Lets you check the price and specs of Cortical Labs’ CL1.
• “World’s first bioprocessor uses 16 human brain organoids”, Tom’s Hardware, 2024. Link ··· Ground-level details on FinalSpark’s organoid rental service.
• “The Human Brain Runs on Less Power than a Light Bulb”, Britannica. Link ··· Lays out both the basis for the 20-watt figure and its limitations.

---

The author, Kwangseob Ahn, is a professor of business administration at Sejong University and lead consultant at OBF (Oswarld Boutique Consulting Firm). He teaches statistics and data analysis, including business data management and business analytics, while leading GTM and AI strategy consulting in the field, designing the seam between technology and business. He has published academic research on a memory architecture for AI dialogue systems (HEMA) and runs Daily Arxiv, a daily curation of global AI papers. He holds a master's from Korea University's Graduate School of Technology Management and a KMBA. He is the author of Homo Brainless: The People Who Outsource Their Thinking.

Footnotes

1. DishBrain: A system made by growing living neurons on an electrode array. When signals were exchanged with it, it showed adaptation to “Pong” within 5 minutes. Whether this can be called “intelligence,” though, is still under debate. ↩

2. Organoid: A very small clump of tissue grown from stem cells in a lab, mimicking an organ. A brain organoid imitates some of the structure and activity of a real brain, but it doesn’t think the way a human brain does. ↩

3. Neuromorphic computing: A chip design that mimics the brain’s structure. Unlike conventional chips that run in lockstep to a central clock, it exchanges signals only when needed, using far less energy on optimization problems. ↩