Because it's obvious to someone who can move out of the shallow end of language. Its a concept that arises from the value of self, which controls behaviour. A machine may or may not have machine learning algorithms, but since both machines lack a self, any distinction between them is vacuous. You have to introduce a sense of self to distinguish these two sentences: A. If the activites are unpredictable, then freewill will be present. B. If the activities are unpredictable, then a lack of control will be present. A is controlled by the self. B is controlled by random generation. You can say that unpredictable action is a necessary condition for free will, but not a sufficient one. Usually the (deterministic) response to this is to define the self in a counterfactual manner eg. "A lack of control in deterministic paradigms gives us unpredictable behaviour. The self also gives us unpredictable behaviour so it either (a) arises from a lack of control or (b) is actually predictable" As such, subscribers to the deterministic notion of freewill commonly point to engineering random generation in machines, which, in their minds, illustrate the relationship between a lack of control (random behaviour) arising from a predictable language (computer programming). Which then leads them to their next adventure in folly : wading in the kiddy pool of language while imagining they are participating in some sort of Jules Verne sequel. If one can actually move between the deep and shallow notions of language one can avoid all this .... hence the "yada, yada".