Introduction. The article analyzes the structure of the epistemology of knowledge of weak artificial intelligence in comparison with the structure of human knowledge. The article was written within the framework of one of the most demanded branch sections of philosophy – the philosophy of artificial intelligence. The novelty of the research is based on the idea of applying the analysis of knowledge developed within the framework of analytical epistemology to the field of artificial intelligence.…
Read moreIntroduction. The article analyzes the structure of the epistemology of knowledge of weak artificial intelligence in comparison with the structure of human knowledge. The article was written within the framework of one of the most demanded branch sections of philosophy – the philosophy of artificial intelligence. The novelty of the research is based on the idea of applying the analysis of knowledge developed within the framework of analytical epistemology to the field of artificial intelligence.Methodology and sources. The article is written in the framework of analytical tradition in philosophy. We use conceptual analysis to analyze the most crucial aspects of knowledge. This method assumes that in order to explain a complex phenomenon, it is necessary to analyze its components. This method formed the basis of an analytical discussion about knowledge in the second half of the 20th century. The article also uses comparative analysis.Results and discussion. People's knowledge is characterized by three essential characteristics – the presence of information-bearing mental states, their reliability, and factuality. We analyzed to what extent the information-carrying internal states of a weak AI correspond to these qualities. The authors concluded that internal states can be considered beliefs if they are weakly interpreted, they can have a high degree of reliability under certain conditions, and they can have factuality if we accept that a weak AI has beliefs.Conclusion. A weak interpretation of the concept of belief allows us to argue that neural networks are capable of having beliefs. A more rigorous interpretation of the concept of belief also includes the requirement to understand the meaning. However, we do not have at our disposal a satisfactory theory of understanding meaning. In this case, the condition of reliability is the only criterion of knowledge to which the functional states of machines can correspond in the case of certain tasks, in connection with which the problem of generality arises especially acutely.
