It's true that the computer requires a precise formulation of what you want it to do, but I'd submit that current programming methods are very far from the so-c

It's true that the computer requires a precise formulation of what you want it to do, but I'd submit that current programming methods are very far from the so-called minimum description length limit, where we'd be telling the computer nothing but what we want it to do. A framework for encapsulating different problem domains would help, as there are far fewer problem domains than there are programs. If anyone seriously thinks that traditional procedural languages (even the more concise ones ones like Python and Ruby) are going to scale for another fifty years, they're on glue. We can't go on like this.