The Possible Irreducibility of Artificial Software Life

I argue that the characterization of reducibility as computability is too weak to be credible. I propose an improved version -- "intentional computability" -- and show that it is indeed a stronger criteria, in that there may be specifications for which there exist programs but where there is no systematic way to build such a program from the specification. This undermines the assumption that just because software life would run on a computer that it must be reducible to a computer program as the result of an intentional plan. This, in turn, opens the way to the possibility that an intentionally irreducible program might arise (e.g. by evolution)