L’intelligenza delle I.A come effetto di senso semio–narrativo nei giochi digitali. Una rivoluzione semiosofica

The question at the heart of this contribution concerns the impossibility for a human agent to distinguish itself from an AI on the basis of its performative behavior exhibited in a video game and the socio‒cultural implications that this impossibility entails. Thus, this is a work that investigates the effect in light of which we ascribe to the AIs present in most common video games the quality of being intelligent. This investigation begins with a semantic observation on the performative implications that are present in the very definition of the notion of "intelligence" in many dictionaries. Implication we point to as an ideological connotation insofar as the intelligence characteristic of the human species seems to be able to be reduced to the narrative demonstration of knowing how to do and doing well. We then turn from dictionaries to social discourses about digital games and observe under what conditions A.I.’s create a sense effect of stupidity and under what conditions they appear to be intelligent. We then focus on the latter and semiotically explain this attribution on the basis of two different ways in which it manifests itself in texts: a narrative‒mimetic construction and a pragmatic one. In the first case we describe a mirror situation between human agent and machinic agent, such that in fact the agency of the A.I. is naturally interpreted in narrative terms through the categories employed by the human agent itself. In the second case, we analyze the different technologies behind the demonstrated effectiveness of A.I. (F.S.M, G.O.A.P, B.N.N., Utility A.I., Reinforcement Learning, Behavior Tree) in some iconic video games (Pac–Man, Half‒Life, Fear, Alien Isolation, Black & White, Forza) and show how these systems are built on a modelization of performative human action viewed from different aspects (effective problem solving, adaptivity, multiplicity of options, contextualization, potentiality, inferential construction of a model opponent, and experience). Thus demonstrating an almost impossible distinction between A.I. and human within a performative framework, we argue in our conclusions that digital games are an important part of contemporary social discourse around the unicum of human intelligence and propose two alternatives to a functional‒performative conception of it: an environmental and a ludosemiotic conception.