Phew! At last, I can warmly recommend another paper in Church’s Thesis after 70 Years …

Although initially Gödel was hesitant, by about 1938 he is praising Turing’s work as establishing the “correct definition” of computability. Yet in 1972 he writes a short paper on undecidability, which includes a section headed “A philosophical error in Turing’s work”, in which Gödel takes issue with Turing’s implicit assumption of the finitude of the human mind. Question: is this a change of view on Gödel’s part (from seeing Turing’s work as successful to seeing it as fatally flawed)?

Surely not. What Gödel was praising in 1938 was Turing’s analysis of a finite step-by-step computational procedure. (Recall the context: Turing was originally fired up by the Entscheidungsproblem, which is precisely the question whether there is a finitistic procedure that can be mechanically applied to decide whether a sentence is a first-order logical theorem. So it is analysis of such procedures that is called for, and that was of concern to Gödel too.) What the later Gödel was resisting in 1972 was Turing’s further thought that, in the final analysis, human mental procedures cannot go beyond such finite mechanical procedures (he was inclined to think that, in the words of his Gibbs Lecture, the human mind “infinitely surpasses the powers of any finite machine”). There is no change of mind, just a change of topic.

That’s at any rate what I have previously taken to be the natural view, without doing any careful homework to support it (but it is of course the view that fits with the important distinction I’ve been stressing in these comments, between CT as a claim about effective computability, and bolder theses about what can be computed by machines and/or minds). And so it is very good to see that now Oron Shagrir has given a careful and convincing defence of this view, with lots of detailed references, in his (freely downloadable!) paper “Gödel on Turing on computability”. Good stuff!