I’ve spent a lot of time focusing on what AI will do to the individual mind. From a student who stops fighting to a professional who doesn’t trust his own judgment, it was eye-opening. And while these are real concerns, there may be something bigger going on.
Here is the main question. What happens when millions of people do the same “cognitive business” at the same time? It is not the failure of any single mind, but what is manifested as a whole.
The difference is the point
In general, complex systems do not self-regulate through consensus. They correct self-regulation through the effective friction of people who see things differently. And when that happens, they come to different conclusions, for reasons they sometimes can’t fully articulate.
Markets are a great example. The financial market, in fact, is not a calculation. It is a mechanism for collecting different human opinions. In any transaction, the buyer and seller disagree about the value. This discrepancy is not a flaw in the system, it is the system. Remove the friction and you don’t get a smarter market. You get a more delicate one.
Science, medicine and markets all depend on the same hidden and very human asset. They are full of people who look at the same fact and come to different conclusions. And it can be very expensive.
Passive convergence
The calculator has expanded our calculation capabilities. A search engine has expanded our access to information. It’s not about how people judge architecture.
AI does. And when that happens, the spiral begins. When the same models, trained on the same data, process the same inputs, inform the risk assessment of investors, the diagnostic evidence of doctors and the editorial opinion of writers, something structural changes occur. Individuals remain. Diversity leaves. And as soon as you leave it, the work that people produce becomes more similar. And this will feed the next generation of models, which will narrow the range even further.
I wrote about cognitive surrender— the way humans delegate judgment to AI, not because it’s hard, but because it’s easy. I wrote about collapse of the modelwhat happens when AI recursively trains on AI-generated data and originality drifts away from the average. Synchronous blindness is both of those things happening at the same time in the population. Not one person surrenders, but everyone gathers.
What is lost in tails
My point here is that statistical tails are important. financial bubbles, intelligence failures, medical assumptions that persisted after the evidence contradicted them. These were not information failures. They were a failure of perception. Enough people looking at the same facts and missing one thing at a time. In the markets, an experienced investor who senses something wrong before the information that it operates on. pattern recognition created through memory about what a bad position is really worth. That knowledge lives where the models don’t. It cannot be mined or encrypted. It must be earned.
When this kind of perception is lost by the number of participants, the system loses its early warning capability. The anomalies that might have found a sufficiently varied field of human judgment are lost. And it’s not because they’re not there, it’s because the device that sensed them is integrated around the middle. No tail, no telling.
Systemic risk we don’t measure
We have a framework and clinical tools for individual measurement cognitive decline. We have very little to measure this narrowing of collective perception. This is a hidden problem because the risk of blindness is not visible at the individual level. Anyone can do better than ever before using the convenient and powerful AI tools. The collection, at the same time, can be seen less.
AI is not only changing the way people think, it is also changing what we can perceive. And when that perception narrows collectively, the danger is not that AI will fail us. This is how we all simultaneously miss important things without knowing it.
