The analysis of cognitive functions of dilemma are undertaken to show the substantial character of thinking and limits of its formalization in AI programs. Substantial features of complex dilemmas are considered, the concept “solution of dilemma” and “resolution of dilemma”, as well as the task of choosing between the dilemma alternatives are analyzed. It states that resolution of dilemma is a formal-logical algorithm of inference, provided by the axiomatic status of dilemmas; the solution of dilemma involves the disclosure the contradictions and searching for new relations between its terms and alternatives. This thesis is proved through the task taken from life experience of G.S. Altshuller (Russian inventor). Distinctive features of moral dilemmas formalization and the meaning of non-logical factors for its solution are discussed. The limitations of programming the morally qualified AI systems implemented in rigoristic ethics and natural law terms are shown.

Keywords: dilemma, contradiction, Artificial Intelligence, thinking

This article examines some aspects of hypothetical interaction between Artificial Intelligence (AI) device and a man in which the goal of AI is self-learning to think. The practical reason of this task is to draw attention to a number of significant features of thinking that are important for AI engineering. The main issues of AI methodology and its critics are analyzed. It discusses the features of a person's thinking, which will inevitably face the AI​​, even though they have marks of formalizing impossibility as well as positive heuristic properties and functionally irremovable out of thinking. Analogy, complexes and pseudo-concept, contradiction are considered among these features. It's shown that communication is a source of new information imparted to the system essential for thinking and ensuring its development.

Keywords: thinking, Artificial Intelligence, analogy, complex thinking, contradiction.