In his paper "Intelligent Machinery" Alan Turing suggested the idea of building a `thinking machine' by making a machine that emulated as many parts as possible of a person, and letting it "roam the countryside" finding things out for itself -a robot that was to learn from its experience in the ordinary world. Turing rejected that idea as not practical at the time and moved on to more disembodied suggestions of how to build a thinking machine. Now that we have experience with humanoid robots we can examine some of his briefly expressed ideas on such robots, and the challenges of what might be missing in such an enterprise, and see how it is playing out now that it is practical to build the machines he suggested. But even further we can re-examine Turing's models of human behavior to produce a formalism for computation, and compare that to computational neuroscience which tries to explain human thought as computation.