Or: As Entropy Increases, Many Weak Cues Dominate What's the difference between software 1.0 and software 2.0? Software 1.0 is when we approach a problem by considering a small number of strong features. When we answer a question by giving a small number of strong reasons. When we write plans by saying a then b then c. And contingency plans by saying if a then b else c. Boolean logic is hard and focused. We never have if statements nested 100 levels deep. Software 2.0 is the opposite. When we use "A.I.", we're choosing to tackle a problem by considering a large number of weak features, rather than a small number of strong ones. To answer a question by giving a large number of inconclusive answers, surrounding and suffocating the question rather than attacking it head on. Drowning the ambiguity with a thousand tiny reasons that each in isolation seem kind of ok, but not great. Therefore, in situations of high entropy, the latter methods will inevitably dominate the former, as they're more robust to any perturbation of the assumptions they require to execute correctly.