This detracts from the point of the article. The author is not writing a program, nor is he aiming for "perfect correlation". He is noting a series of observations and proposing some connections between them, as is often the case with writings about society. If we take a level of rigor that works for sub-100KLOC algorithms running on dozens of cores and try to apply it to millions of interacting neural networks, each with 100 billion neurons, there will hardly be any writings about society.