Can AI Really Understand? Semantic Limitations and the Boundaries of Algorithmic Cognition

This paper undertakes an interdisciplinary analysis of the problem of understanding and knowledge in the context of contemporary artificial intelligence models. The starting point is the distinction between the weak and strong versions of the AI hypothesis, which reflect two fundamentally different visions of machines’ cognitive potential. The article focuses on the question of whether algorithmic information processing can lead to genuine semantic and conceptual understanding. Against the backdrop of classical epistemological theories and the philosophy of language, the paper examines the limitations of the syntactic approach to cognition, in which knowledge is reduced to operations on symbols devoid of reference to the world or meaning. Special attention is given to John Searle’s “Chinese Room” argument, which illustrates the fundamental difference between formal symbol manipulation and conscious understanding. The author argues that while artificial intelligence can efficiently perform complex cognitive tasks, it lacks intentionality and the capacity to transcend the symbolic domain.