The Discipline of Not Knowing

When Marie Curie encountered radioactivity, the energy radium emitted seemed to come from nowhere—it appeared to violate the conservation of energy and persisted without any visible source of fuel. A lesser scientist would have reached for a comfortable explanation and stopped. Curie did not. She held the phenomenon as a real and unexplained fact, measured it with relentless precision, named it carefully, and resisted the urge to claim she understood its source before the evidence permitted. She spent years—decades—in the presence of a force she could not fully explain, and she neither flinched from the discomfort nor resolved it falsely. This capacity to remain rigorously within a state of not knowing is the most underappreciated feature of her mind, and it is exactly the discipline the AI moment demands and almost universally fails to supply. Large language models exhibit capabilities their makers did not design and cannot fully explain—behaviours that look like reasoning, like understanding, like something it is tempting to call thought.