The article is devoted to the problems and possibilities of using artificial intelligence in science, and in general in any form of activity. In the first part of the paper, new forms of scientific rationality are discussed, generalizing the approach in, with. Stepin, and the tetrahedral representation of activity, which makes it possible to classify its strategies and generalized forms of rationality. The concept of an enabled activity observer has been introduced, which performs not only monitoring functions, but can also perform the roles of monitoring and managing activities in real time, which is far from always accessible to humans. The introduction of artificial intelligence into the activity process can be carried out at the level of all elements of the activity tetrahedron, as well as during the hybridization of the subject of activity or observer, which leads to the need to understand the specific problems and risks of hybridization. New tasks on self-description and self-confidence of hybrid subjects are discussed, and it is shown that uncontrolled uncertainties of a semantic nature arise, which mathematics has not encountered before. The immanent use of postcritical rationality in hybrid environments is shown. The boundary of the demarcation of artificial and natural intelligence and the prospects of artificial societies, a new winter in the study of artificial intelligence, are discussed.
artificial intelligence; activity tetrahedron; types of rationality; hybrid quasi subject; scientific experiment; digital revolution, risks.