On Artificial Intelligence and Human Stupidity

There is a tiny book written in 1976 that explains more about what is happening inside technology companies today than every AI keynote of the past two years combined.

It was written by Carlo Maria Cipolla, an Italian economic historian who divided his life between Pavia and Berkeley. Cipolla spent decades studying how economies rise and fall, how currencies collapse, and how empires manage to ruin perfectly good things. Eventually, he arrived at a conclusion so liberating—and so politically incorrect—that at first he only printed the book in English for a small circle of friends.

You don’t need dark conspiracies to explain why the world is a mess, Cipolla argued. You don’t even need malice. You just need to understand the single most powerful force in human history: stupidity.

Now, before anyone gets offended, Cipolla wasn't using the word as an insult. He was using it the way a physicist talks about gravity. And half a century later, we have just done something Cipolla never lived to see: we handed human stupidity an infinite amplification engine called Artificial Intelligence.

Cipolla’s Five Laws (Briefly and Precisely)

Cipolla’s five laws take less than two minutes to learn, and once you know them, they explain almost everything you see at work.

1. The First Law: Always and inevitably, everyone underestimates the number of stupid individuals in circulation. No matter how high your estimate is, you are still low. People you once assumed were rational and level-headed will, at the most inconvenient possible moment, surprise you.
2. The Second Law: The probability that a certain person is stupid is completely independent of any other characteristic of that person. This is the law that bruises the most egos. Whether you test janitors, middle managers, Silicon Valley founders, four-star generals, or Nobel Prize winners, the exact same fraction of people in every group is stupid. Credentials, wealth, and fancy titles don't reduce stupidity; they just give it a bigger megaphone.
3. The Third (and Golden) Law: A stupid person is a person who causes losses to another person or to a group of persons while deriving no gain for himself, and even possibly incurring losses.
4. The Fourth Law: Non-stupid people always underestimate the damaging power of stupid individuals. In particular, we keep forgetting that at all times, in all places, and under any circumstances, dealing or associating with stupid people always turns out to be a costly mistake.
5. The Fifth Law: A stupid person is the most dangerous type of person—even more dangerous than a bandit.

To show why the Third and Fifth Laws are true, Cipolla drew a simple coordinate plane measuring every human action by two questions: How much does it help or hurt you? and How much does it help or hurt others?

FIG. 01 · CARLO MARIA CIPOLLA’S FOUR QUADRANTS OF HUMAN ACTION (1976)HOVER OR TAP ANY QUADRANT TO INSPECT ITS COORDINATES

Look at the four quadrants that emerge:

If you do something that benefits you and benefits the people around you, you sit in the top-right corner: The Intelligent (Gli Intelligenti). You are playing a positive-sum game. You leave the room better than you found it.

If you do something that hurts you while creating a gain for someone else, you are in the top-left corner: The Helpless (Gli Sprovveduti)—generous, perhaps, or simply naive.

If you do something that benefits you by inflicting a loss on someone else, you sit in the bottom-right corner: The Bandit (I Banditi). Now, nobody likes a bandit. If a bandit mugs you in an alley and takes a hundred dollars, you are a hundred dollars poorer. But the bandit is a hundred dollars richer. There is a logic to it. You can predict what a bandit wants, you can lock your doors against a bandit, and you can even negotiate with a bandit.

And then there is the bottom-left corner: The Stupid (Gli Stupidi).

Why is a stupid person more dangerous than a bandit? Because a bandit at least follows rational self-interest. A stupid person will walk into your kitchen at two in the morning, set your toaster on fire for no reason at all, burn his own eyebrows off in the process, and walk out complaining about the smoke. Everyone loses. Nothing is gained. And because there is no rational motive, you can never defend against it.

How Corporate Boxes Manufacture Stupidity

If you have ever spent time inside a corporation, you know the cruel irony of Cipolla’s diagram: leadership teams love putting human beings into four boxes, yet they almost never use the right ones.

There is an entire genre of internet memes dedicated to corporate quadrants. Lock ten executives in an off-site room with a whiteboard and a plate of croissants, and within twenty minutes someone will draw two perpendicular lines, invent two completely made-up axes—“Strategic Velocity” versus “Cross-Functional Alignment”—and spend the next three quarters shoving every human in the building into a square.

Almost all of those corporate grids are pure theater. In my entire career, aside from Cipolla’s chart, I have only ever seen one other two-by-two chart that actually tells the truth about how organizations work.

Years ago, Simon Sinek gave a famous talk about how the Navy SEALs choose operators for SEAL Team Six. They draw a simple chart with Performance on the vertical axis and Trust on the horizontal axis.

FIG. 02 · SIMON SINEK / U.S. NAVY SEALS PERFORMANCE VS. TRUST MATRIXHOVER OR TAP ANY QUADRANT TO INSPECT ITS TEAM IMPACT

Obviously, nobody wants the person in the bottom-left with low performance and low trust, and everybody wants the star in the top-right with high performance and high trust. What fascinated Sinek was what the SEALs discovered about the other two corners: they will take a medium—or even a low—performer of high trust over a high performer of low trust every single time.

Why? Because in a real team, the “high performer of low trust” is poison. He is the superstar who hits his personal targets by stepping on his teammates' necks, hoarding credit, and letting everyone else take the fall. Look closely at Sinek’s chart and you realize that a high-performer of low trust has a much older name in Cipolla’s book: he is a Corporate Bandit.

Most companies are terrible at telling the difference between an Intelligent builder and a charming Bandit. And because leadership teams don't know how to measure trust or judgment, they try to protect themselves by inventing Processes.

And that is where institutional stupidity is born.

Take one of the most surreal chapters in corporate history. For years, Microsoft ran its performance reviews on a rigid “stack ranking” system. Under this process, every manager was forced by default to grade their team on a fixed statistical curve—meaning that every review cycle, a set percentage of every single team had to be placed at the bottom and managed out the door, on the mathematical assumption that a fixed percentage of hires must be wrong hires.

Pause for a second and admire the sheer architecture of that idea.

You could be a manager who spent two years recruiting ten of the brightest engineers on earth—ten people who belonged in the hall of fame. Didn't matter. The spreadsheet demanded a sacrifice, so one of your geniuses had to be labeled a failure.

What happened next? Exactly what any sane person would predict. Smart engineers refused to join teams that already had smart engineers, because sitting next to a genius meant you might be the one pushed off the boat. People stopped helping other teams. Colleagues hoarded information so their peers wouldn't outshine them. An entire decade of innovation was strangled by an HR formula.

Ask yourself Cipolla’s question: Who gained from that process?
Not the engineers. Not the managers. Not the customers. Not the shareholders. It was pure, industrial-grade bottom-left stupidity—dressed up as “rigorous management science.”

Enter AI: Removing the Speed Limit on Bad Ideas

For the last few thousand years, humanity had one built-in safety mechanism against Cipolla’s Fourth Law: it took effort to do things.

Before AI, stupidity had a natural speed limit. Even the most tirelessly counterproductive person in your company only had ten fingers and an eight-hour workday. There were only so many five-page memos they could write, only so many unnecessary reorganization decks they could format, and only so many people they could distract before Friday afternoon.

Then we invented generative AI—and took the brakes off the car.

Across the business world, people are doing pointless things at ten thousand times their previous speed.

Someone has a tiny, half-formed thought on a Tuesday morning. Instead of thinking it through for ten minutes, they ask an AI to turn three bullet points into a majestic, twelve-page “Strategic Vision Document.” They send it to twenty coworkers. Those twenty coworkers take one look at the twelve pages, roll their eyes, and ask their AI to summarize it back into three bullet points.

Think about the physics of what just happened. Twenty people lost twenty minutes of their day, a data center somewhere in Virginia boiled a swimming pool of water to run the GPUs, and not a single drop of real human insight moved from one brain to another.

And that is just the polite version. Across the internet, companies are unleashing armies of “autonomous AI outreach agents” to flood millions of inboxes with fake-personal messages—torching their own reputation while making email unusable for everyone else. Product teams are generating mountains of features nobody asked for and code nobody understands, simply because the button to generate it is free.

The Great Boardroom Delusion

The single biggest delusion in corporate boardrooms right now is the belief that plugging AI into your existing process is going to make your company smart.

Everywhere you go, executives are panicking: “Put AI into our customer support! Put AI into our planning process! Put AI into our performance reviews!”

If your process is already broken—if your support flow is designed to trap customers in a maze so they give up, if your product process is nineteen layers of committees approving slide decks, if your culture rewards Corporate Bandits and punishes honesty—plugging AI into that process will not fix your company.

It will just make your stupid process stupider, at the speed of light.

Imagine if Microsoft, at the peak of its stack-ranking madness, had plugged an AI agent into HR so it could stack-rank and fire a percentage of every engineering team automatically every Monday morning. Would that have been “digital transformation”? Or would it have been a faster way to burn the house down?

Technology has never once fixed a broken philosophy.

Why I Am Still an Optimist: AI in the Hands of the Intelligenti

Now, you might read all of this and think I am cynical about AI.

Quite the opposite. Once you understand Cipolla’s laws, you realize that blaming AI for the flood of corporate nonsense is like blaming a hammer because someone hit their own thumb.

AI is a tool.

A tool has no ego, no morality, and no quadrant of its own. It is a pure multiplier of the human being holding it. Hand it to a Bandit, and they will use it to extract more while giving less. Hand it to a stupid process, and it will manufacture chaos on an industrial scale.

But hand that exact same tool to someone in Cipolla’s Intelligent quadrant—someone non-stupid, with a non-stupid purpose—and magic happens.

Remember what defines the Intelligenti in Cipolla’s world, and what defines the top-right corner of Sinek’s Navy SEALs chart: high trust, clear judgment, and actions that create real value for yourself by creating real value for others.

When people like that pick up AI, they don't use it to generate twelve-page memos nobody wants to read, and they don't bolt it onto yesterday's bureaucracy. They start from first principles. They ask how something should actually work for another human being. They strip away the silly boxes, the committee theater, and the artificial friction. And then they use AI as a lever—allowing a small group of high-trust builders, or even a single clear-minded creator, to build things of extraordinary usefulness and beauty for the world.

In the end, the age of AI isn't really a test of machine intelligence at all. The machines are plenty smart enough for what we ask of them.

It is a test of ours.

You can use the most powerful tool in human history to automate your bad habits and multiply the noise. Or you can step into Cipolla’s top-right corner—fix how things work, care about the human on the other side, and build something that leaves everyone in the room better off than before.