As máquinas e a língua : um debate entre a Inteligência Artificial de Turing e a Enunciação de Benveniste

Abstract

The intersection between Artificial Intelligence (AI) and Linguistics is more than fundamental, it is necessary. The concept of the Turing Test is proposed since the early days of the modern computer science, which corroborates this intersection, equating, in potentiality, machine and human. This work aims, from this Turing conception, to oppose the proposal with linguistic aspects abstracted from the reflections on language in Émile Benveniste, especially from a distinction path between the semiotic and semantic domains to reach the enunciative conception. This theoretical movement helps to understand what Searle presents as a semantic problem of theories based on a strong Artificial Intelligence. The objective, therefore, is to understand what the limits of language manipulation by intelligent agents are and what subsidies the own research can offer to our understanding of human language. The concept elaborated by Turing and the development of AI in contemporary times, well described by Stuart Russell and Peter Norvig, already guarantees a genuine potential in a linguistic scope for AI. However, this aspect depends especially on a selection and combination of terms by probability of use, without especially reaching a human character of the language, which is explained by Searle, in the argument of the Chinese Room and can be understood through the concept of language semantization, necessary for enunciation, abstracted from Émile Benveniste. It is precisely because of this condition that human beings distance themselves from other animals and also from machines. The ability to combine, associate and define probable terms is extremely useful in some conditions with a well-defined environment, because in these protocol situations it is interesting the rapid diagnosis that the machines are capable of producing, as in a screening of medical symptoms, for example. This does not occur in everyday situations, in which the language is updated with each proposed enunciation. Conversation in a pragmatic context can occur and ensure that the Turing test is passed, without the machines being capable of discursive production, an element that comes from the enunciative process, internal, personal and unique. The language is updated with each enunciation, and this is a condition that affects the meaning through the sense, this becomes a problem for any intelligent agent that processes language, making machines dependent on humans for the maintenance of natural and living language. The relation between man and machine can be seen, therefore, from another perspective, of complementarity, while one guarantees the process of updating the language, and the other assists with the sorting of information in a digital universe immeasurable by the human being.