THE UNDEFINED EVENT : On "AGI," the Probability Attached to It, and Why a Forecast Without a Specification Cannot Be Wrong

A probability is a claim about the likelihood of an event. The event must be defined — notprecisely, but enough that a reasonable person could say whether it occurred. "AGI" does not meet that standard. It has no agreed definition. It has several competingdefinitions, each of which implies a different event. And each event would carry a differentprobability. Musk's reported 10% is therefore not a probability. It is a hedge: a number attached to an outcome that has not been specified, and therefore cannot be checked. If a definition is ever agreed, the 10% can be reinterpreted. If one never is, the number was never wrong. This is not a semantic quibble. It is the mechanism by which AGI forecasts survive contact with the future. An undefined event cannot be refuted. A forecast that cannot be refuted is not a forecast. The paper states what a defined event would look like, why the current usage fails the test,and what changes when the word is replaced by a specification.