Accepted_test

Programming Methodology for Trusted Artificial Intelligence
by Sergey Goncharov | Andrey Nechesov | Dmitry Sviridenko | Sobolev institute of Mathematics, Novosibirsk, Russia | Sobolev institute of Mathematics, Novosibirsk, Russia | Institute of Philosophy and Law, Novosibirsk, Russia
Abstract ID: 244
Event: BGRS-abstracts
Sections: [Sym 12] Section “Systems theory, big biological data analysis, ontologies and artificial intelligence”

The task of building reliable, transparent and explainable artificial intelligence attracts the attention of all the world's researchers. Today it becomes clear that all advanced large language models do not satisfy even the basic requirements put forward for trusted intelligent systems.
The report will touch upon the development of trusted artificial intelligence algorithms.  A programming methodology in Turing complete languages will be presented, which is based on the task approach and the concept of semantic programming.
Our task is to make the thinking process of intelligent machines transparent and human-understandable. For these purposes, it is also necessary to build a coherent and high-quality concept of learning theory.  This concept should allow, based on logical rules and basic statements, to show the user the final result produced by the intelligent system. For these purposes, a new logical-probabilistic method will be presented, which allows us to close all these gaps.