CONCEPT
Organized Complexity
Weaver's 1948 name for the great unexplored middle between tidy simplicity and pure randomness—problems with a sizable number of variables “interrelated into an organic whole”—which turns out to be the exact territory machine learning was built to conquer.
Warren Weaver drew a map of all scientific problems in a 1948 essay called “Science and Complexity,” and at its center he placed a territory that neither classical method nor statistical aggregation could reach. Problems of
simplicity involve a handful of variables and yield to equations; problems of
disorganized complexity involve millions of independent variables and yield to statistical averages. But organized complexity is the middle kingdom: problems with “a sizable number of factors which are interrelated into an organic whole,” where the variables are too many for equations and too
coupled for statistics to capture the relationships. Language, biology, economics, city life, and mind all live here. Weaver predicted in 1948 that conquering this territory would be the great scientific labor of the coming half-century and would require computers “programmed more like a brain.” The prediction came exactly true: every flagship achievement of
large language models—language, image recognition, protein folding, strategic reasoning—is a conquest of organized complexity,