Abstract: In this paper, the authors present their original concept of explainable and trustworthy artificial intelligence (AI), which is based on a task approach to artificial intelligence. The purpose of this work is to show that the task approach to AI generalizes and develops such well-known approaches as the agent-based approach to AI and artificial general intelligence (AGI). In addition, unlike AGI, it is anthropomorphic, since, it formalizes the physiological Theory of Functional Systems (FTS) of brain function [15-20], which describes purposeful human activity. For the purpose of effective practical implementation of the task approach, the work presents variants of executable task specification languages that allow, based on the explanatory and interpretable specification of tasks, to generate their executable code leading to solutions. Specifications of tasks related to a certain subject area, for which a corresponding query to the subject area is generated, the search for an answer to the query and obtaining a solution (answer to the query) is carried out within the framework of the semantic modeling developed and presented in the work. Using various specification languages, several platforms for solving applied problems have been developed and presented, which, together with examples of solved problems, are presented in the conclusion of the work.

Cite: Vityaev E.E. , Goncharov S.S. , Gumirov V.S. , Mantsivoda A.V. , Nechesov A.V. , Sviridenko D.I.
Task approach: on the way to trusting artificial intelligence
In compilation ИЗБРАННЫЕ ТРУДЫ. Всемирный конгресс. Теория систем, алгебраическая биология, искусственный интеллект: математические основы и приложения (26–30 июня 2023 г., Москва).. 2023. – ISBN 978-5-91291-065-4. DOI: 10.18699/sblai2023-41 OpenAlex

Dates:

Published online:

Nov 1, 2023

Identifiers:

OpenAlex:

W4389684239

Citing: Пока нет цитирований

Altmetrics: