I retract my claim. Satoshi Nakamoto is definitely a human(or trying to act like one). A genius at that. He made two mistakes in his posts: walkthough and idealogical. So what if he had invented an A.I. with which he created a bitcoin. He would have obviously made mistakes or made inelegant code just to throw us off. How about that?
With this hypothesis, we can safely say self recursively improving A.I. existed since 2008/2009. But we have not found a way to make it. Whereas Satoshi made it and used it on his own accord. But then the question again is "Who is Satoshi?"
There is no problem to create a self learning AI
The theories behind machine learning had been properly developed some 40-50 years ago. The main problem is that it is often not worth doing it. In most real world cases you will get better results with less costs by logic being hard coded (explicit programming), i.e. either by directly employing some quantitative theory already existing in the field of your interest or by using statistical approach (when you can't rely on such a theory)