Правовая квалификация модели машинного обучения на основе технического анализа этого процесса

Abstract

Machine learning technologies are at the heart of artificial intelligence applications, ranging from use in the military to household assistants. The result of machine learning as a technical process is the creation of models that are able to produce a forecast of some degree of accuracy, on the basis of which people make decisions, and artificial intelligence systems perform actions. The machine learning model is a new phenomenon with an indefinite legal regime, however, regarding the creation and use of models, legal relations are formed, contracts are concluded, economic benefits are created during machine learning: all this needs legal qualification. Since inductive development methods are used to create models, machine learning has an atypical structure compared to a conventional computer program, which makes it difficult to determine the legal regime of the model and the legal consequences of its creation and operational application. The article presents an element-by-element technical analysis of machine learning as a process and a model as its result. On this basis the legal nature of the model is determined at all stages of its life cycle, protected elements of the structure are indicated; the legal significance of machine learning methods for the legal qualification of models is analysed, the character of training data external to the model is substantiated, and a benefit derived from training data in the form of model parameters is indicated. The conducted research is the basis for further analysis of legal relations that develop regarding the creation and use of artificial intelligence applications.