On the basis of analysis of history of scientific discoveries, and also data of cognitive psychology and theory of artificial intellect it is established that the main factors excluding algorithmic character of creative activity in science are the probabilistic nature of inductive conclusion, the method of trial and error (method of sequential search), the factor of case in scientific discovery and Gedel's theorem of incompleteness. This allows us to solve the 18th problem of S. Smeil: what are the limits of intelligence, both artificial and human?