On "Turing Machines" and Artificial Intelligence

In his recent book, "Turing's Cathedral", historian of technology George Dyson (son of Freeman and sister of Esther) says the following: "The paradox of artificial intelligence is that any system simple enough to be understandable is not complicated enough to behave intelligently, and any system complicated enough to behave intelligently is not simple enough to understand. The path to artificial intelligence, suggested Turing, is to construct a machine with the curiosity of a child and let intelligence evolve".  Leaving aside our advantage of being able to look backwards in time at Turing and his machines, is that observation true? How did artificial intelligence evolve from Turing's "simple" calculating machines to its present state? Does machine "consciousness" have to rise to the level of Ray Kurzweil's "singularity" for this statement to be true? I think the answer is no for several reasons. First, we already have incredibly complex systems which, while to a degree learning-recursive, do not fully act "intelligently". Second, isn't general relativity a system which behaves intelligently but is capable of being understood by us? Third, and I pose this as a question to coders - are we not still far from the point of true AI?  Yes, we have recursive learning. Yes, we have machines that can "write" poetry. Is that enough? We humans are incredible electrochemical and electromechanical machines. Will our functionality (not our genetic material) be duplicatable in a machine, even if every electrochemical/mechanical process in our bodies is finally understood?  Will the machines that artificially run or duplicate those processes function well when randomness is factored in? I think not - I don't think machines will have the curiosity of children and, as one who views the "singularity" as  a dystopia, I hope that "type" of intelligence does not evolve.