Artificial Intelligence under the Scrutiny of Theories of Knowledge

Abstract

This article discusses whether “knowledge” and “intelligence” can be attributed to machines in light of major theories of knowledge. Drawing on Hume’s empiricism, the rationalism of Plato and Leibniz, and Kant’s synthesis, it examines the functioning of machine learning models and, in particular, large language models (LLMs), emphasizing the role of repetition, induction, and habit. It then analyzes whether such models can be regarded as genuine bearers of knowledge or merely as sophisticated simulators of human cognitive capacities. The text also addresses the biases that emerge from Artificial Intelligence (AI) training processes and the risk of amplifying pre-existing asymmetries in the underlying data. In conclusion, based on the Theories of Knowledge, it becomes possible to conclude whether cognitive activity can be attributed exclusively to the human subject or whether it can be extended to the machine.