Self, Activity and Culture in the Era of Generative Models. On the 95th Anniversary of F.T. Mikhailov

The article places the philosophical legacy of Felix Trofimovich Mikhailov in the context of the history of cognitive science and artificial intelligence, the debates surrounding it, as well as in the context of modern generative cogni­tive science. It shows that the debates of the 1960s around the possibilities of ar­tificial intelligence modeling belonged to the period of the “good old-fashioned”, i.e. symbolic AI. Since then, we have seen a few of cognitive paradigms shifting, sequentially re-defining approaches both in psychology and in AI. The author examines the interaction between the subject, activity and culture in the context of modern cognitive approaches capable of providing a rigorous explanation of the origin of sociality, communicative and moral norms. The article shows how the concept of distributed generative models can assimilate the legacy of the activity approach by operationalizing the concepts of activity and culture. It is attainable using the concepts of predictive processing and active inference, based on the idea of cognition and activity as varieties of hierarchical Bayesian inference. The author shows that the comprehensive application of these con­cepts can overcome the contradiction between the activity and information ap­proaches formed in Soviet philosophy.