Реферат: Today the problem of reliability and security is very acute in artificial intelligence algorithms. Almost all existing AI algorithms work in centralized environments. This means that at any moment whoever has access to these algorithms can change both the algorithms themselves, and the information and rules with which they operate. As well as for smart contracts in decentralized environments there is an audit problem for AI algorithms. Very often the code that is implemented requires serious competencies to verify and is not always amenable to careful analysis. The code developer (intentionally or unintentionally) can leave holes in security, which can lead to irreversible consequences. In addition to these two problems, there is another important problem in AI: the AI Black Box Problem. When the intelligent device is a "black box" and the input signals give only the result, but this result is not explained. All three of these issues have a huge impact on how people can trust artificial intelligence.

Библиографическая ссылка: Nechesov A.
About three problems of Trusted ARTIFICIAL INTELLIGENCE
WORLD CONGRESS SYSTEMS THEORY, ALGEBRAIC BIOLOGY, ARTIFICIAL INTELLIGENCE: Mathematical Foundations and Applications 26-30 Jun 2023