IQ Got Automated. Wisdom Didn't.

A frontier model just scored in the 99th percentile on the LSAT, cleared the USMLE, and passed competitive programming benchmarks that would have been the bragging rights of a prodigy a decade ago. The headlines framed it as intelligence catching up to humans.

I think we are measuring the wrong thing.

Every one of those benchmarks is a closed problem with a known answer. That is what IQ tests, and their professional cousins, have always measured — the ability to solve well-defined problems inside well-defined rules. For a hundred and fifty years this has been the dominant definition of intelligence in the West, and most of our institutions were built around selecting for it.

The machines have now caught up to it. And the thing that was supposed to be the ceiling of human cognition turns out to be the easiest part to automate.

Two kinds of intelligence

There have always been two.

The first is IQ — raw analytical horsepower, the ability to manipulate symbols, spot patterns, solve puzzles, hold a logic chain in your head. This is what schools test. This is what SATs and GMATs and corporate assessments select for. This is what we mean, most of the time, when we say someone is smart.

The second is wisdom — the ability to decide which problems are worth solving, for whom, under what constraints, and at what cost. It is knowing when the question as asked is the wrong question. It is recognizing that reasonable people hold incompatible values and that the right answer often depends on a context you cannot fully articulate. It is the tolerance for uncertainty when the data runs out.

These are not the same capacity. They are not even close.

Aristotle made this distinction 2,400 years ago in the Nicomachean Ethics, separating sophia — theoretical knowledge of eternal truths — from phronesis — practical wisdom, the discernment of right action in particular circumstances. He argued phronesis could not be taught algorithmically. Only cultivated, through experience and reflection.

Modern psychology rediscovered the distinction and measured it. Paul Baltes and Ursula Staudinger at the Max Planck Institute spent two decades operationalising wisdom through what they called the Berlin Wisdom Paradigm. Their finding: IQ and wisdom are largely uncorrelated. High cognitive ability barely predicts wise judgment. Age, life experience, and mentorship predict it far better. IQ peaks in your twenties. Wisdom, if you cultivate it, peaks in your sixties.

Igor Grossmann at Waterloo pushed it further. In longitudinal studies, wise reasoning predicted life satisfaction, relationship quality, and emotional wellbeing significantly better than IQ. In one 20-year study, higher wise-reasoning scores correlated with lower mortality, independent of intelligence or education. Wisdom literally predicts who lives longer.

Robert Sternberg, who spent his career at Yale and Cornell working on this, puts it plainly: "We select future leaders for their ability to score well on tests of analytical intelligence, then we are surprised when they lead badly."

How we forgot

We did not forget accidentally. We forgot structurally.

The industrial economy needed reliable solvers of well-defined problems at scale. Filling factories, running bureaucracies, engineering bridges, processing forms, writing software to specification. IQ is a reasonable proxy for all of that. So we built schools, credentials, and hiring systems around it. We compressed a narrow slice of cognition into a number and let that number decide careers.

It worked. For a while.

The Asian education model took this logic the furthest. The gaokao in China, the suneung in Korea, the Japanese entrance exams — high-stakes one-shot tests of narrow cognitive performance that still decide the trajectory of tens of millions of lives. The irony is that Confucius, whose civilization these systems claim as heritage, distinguished zhì — practical wisdom, the capacity to know what is right and act on it — from mere cleverness or book-learning. The junzi, the exemplary person, was not the highest scorer. He was the one who integrated knowledge with virtue and situational judgment.

The imperial examination system began the drift toward memorization and formal mastery. Modern testing culture completed it. The civilizations that most carefully articulated what wisdom is built the institutions that most systematically select against it.

The West did its own version, just with fewer essays and more multiple-choice. Stanford-Binet, Army Alpha, the SAT, corporate IQ proxies, management consulting case interviews — a hundred-year project to find the smartest-on-paper and put them in charge.

And then the Flynn effect, the most uncomfortable data point in this whole story. IQ scores rose roughly three points per decade across the twentieth century. By any ordinary standard, we got much smarter. Yet by almost any measure of wise judgment — long-term financial planning, political polarization, ecological decision-making, institutional trust — we did not get wiser. James Flynn himself acknowledged the tension. The rise was in abstract, decontextualized reasoning. The thing that makes a life go well was untouched.

We got smarter and not wiser, simultaneously, for a hundred years.

What AI actually automates

Here is where this gets interesting.

Large language models are, functionally, enormous pattern-matching engines. They are exceptionally good at tasks with clear inputs, clear rules, and clear evaluation — exactly the profile of IQ-style problems. They can draft code, summarise documents, solve logic puzzles, pass exams. Stuart Russell frames it cleanly in Human Compatible: AI systems optimise for specified objectives and become extraordinarily powerful at that narrow task. What they cannot do is determine which objectives are worth optimising for in the first place.

AI is IQ automation. It is not wisdom automation.

It does not bear the consequences of its recommendations. It does not hold plural, incommensurable human values in tension. It does not recognise when the problem as specified is the wrong problem. It has no stake in what happens after the output ships. Nassim Taleb has been making a version of this argument for years: intelligence is the ability to solve problems, wisdom is knowing which problems are worth solving. The Intellectual Yet Idiot — high credentials, high IQ, no skin in the game — is not a new phenomenon. AI is that phenomenon at industrial scale, without even the pretense of a self to be humbled.

Yuval Harari put it starkly: AI will make human beings richer in data and poorer in wisdom, unless we deliberately cultivate the latter.

The bottleneck has shifted. For two centuries it was computation. It stopped being computation sometime in the last three years. The bottleneck now is judgment.

The new scarce resource

I think we are at the end of the era in which IQ was the primary edge. Not because IQ stopped being useful — it is still the baseline, and always will be — but because it stopped being scarce. When a capability is available on tap, in unlimited quantity, at near-zero marginal cost, it ceases to be what separates winners from losers. It becomes the floor.

What sits above the floor is wisdom.

The ability to look at a benchmark and ask whether the benchmark measures anything that matters. The ability to let a correct answer go because it serves the wrong goal. The ability to hold a decision open in the presence of data that pushes you toward premature confidence. The ability to tell the difference between an optimisation and a trade-off. The ability to know when to trust a machine, when to trust a heuristic, and when to admit that nobody in the room has enough information to decide.

Those who will do well in this era are not the ones with the most IQ-adjacent horsepower. AI will always have more of that. They are the ones who bring something AI structurally cannot: phronesis. Judgment. Context. The willingness to say this is not the problem we should be solving, and be right about it often enough to matter.

This is the shift I keep coming back to. Production got commoditised. Analysis got commoditised. What did not get commoditised — what cannot get commoditised, because it depends on living a life with consequences — is the capacity to know what is worth doing.

The quiet correction

The correction will be slow, because institutions move slowly. Schools will keep testing IQ. Hiring processes will keep selecting for it. Credentials will keep rewarding it. For a while.

But individually, and then in teams, and then in companies, the people who notice this shift first will quietly stop overpaying for the thing that got cheap and start overpaying for the thing that got scarce. They will hire differently. They will promote differently. They will decide differently. They will win differently.

The old game was about being the smartest person in the room. That game is over, because the smartest entity in the room is no longer a person.

The new game is about being the wisest.

IQ got automated. Wisdom didn't.