Eugene Wigner was one of the defining theoretical physicists of the twentieth century. He received the Nobel Prize in Physics in 1963 for his work on quantum mechanics, and above all he lived in the same era as the giants who built modern physics. Max Planck, Heisenberg, Paul Dirac, Leó Szilárd, Edward Teller, Albert Einstein, and John von Neumann. Names that belong in history books were his friends and colleagues.

In his memoir The Recollections of Eugene P. Wigner, he left an interesting assessment of those colleagues. Over a lifetime he had met countless geniuses, but the fastest and most precise mind, he said, was John von Neumann’s.

The mind Wigner describes is close to a near-perfect calculating machine. The moment von Neumann saw a problem he grasped its entire structure, and he could quickly dig down to its most intricate details. He almost never forgot what he had once understood, and he could hold many concepts in his head at the same time. Wigner compared his brain to gears meshing with no more than a thousandth of an inch of error.

But the assessment of Einstein that follows is even more interesting. Wigner says Einstein was slower at calculation than von Neumann and that his memory was not perfect. He had little interest in the details of physics. Instead, what he most delighted in was discovering the most fundamental principles of the world and giving them expression.

Wigner compares Einstein and von Neumann this way:

Einstein took a special pleasure in invention. Two of his greatest inventions are his theories of special and general relativity; and for all of von Neumann’s brilliance, he never produced anything as original — anything no other physicist of the era managed to produce.

This assessment shows that intelligence has different directions. One is the ability to solve an already-existing problem faster and more accurately than anyone else. The other is the ability to redefine, from the start, which problem is even worth solving.

When I look at today’s large language models, von Neumann naturally comes to mind. They connect vast bodies of knowledge, analyze complex problems quickly, and perform in seconds work that would take a human days. Within an existing framework of knowledge, they are astonishingly powerful.

Yet this may also be the fundamental limit of an LLM. An LLM learns from data that already exists and models the patterns within it. It performs brilliantly inside the structure of accumulated knowledge and articulated thought, but doubting that structure itself and proposing a new framework is comparatively hard. I can’t guarantee that future AI will never escape this limit, but today’s LLMs look more like the finest interpreters of existing knowledge.

The Einsteinian ability, by contrast, does not come from the sheer speed of intelligence or the volume of information. It is closer to the ability to see the world differently, to doubt the very framing of an existing problem, and to grasp a principle that has not yet been named.

Einstein’s originality was possible because he rethought the very premises that the physicists of his day shared. Special relativity was not a triumph of computational power but a shift in how we look at time and space. General relativity, too, was born from interpreting gravity not as a force but as the geometry of spacetime.

A new principle is not discovered by applying the existing rules more skillfully. It begins the moment you realize that the existing rules alone cannot fully explain the world. We tend to measure intelligence by problems that already exist in the world. But looking back through history, the people who changed their era were not those who solved problems whose answers already existed — they were those who questioned the problems themselves.

Someday future AI may come to possess that ability too. If so, the important role left to humans now may not be to find answers faster, but to discover what no one has yet thought to ask.

This seems true not only in physics but in art as well. The great artists who endure in history were, for the most part, not those who reproduced an existing style most perfectly, but those who proposed a way no one had tried before. Impressionism, Cubism, abstract art — all were first criticized for breaking the existing rules. Only with time did people realize that they had been new questions.

So it makes me wonder. What is the unknown question we have not yet discovered?