Reliability of algorithms and AI-powered systems and trust in results generated

Abstract

Recently, Artificial intelligence (AI) has attracted a lot of media coverage, much of which is dedicated to biased judgments, "AIdiscrimination," "AI-racism," and other errors identified, which could harm humans and society. Given that, how can enterprises integrate AI into their underlying businesses, if people distrust it works flawlessly? This thesis is aimed to demystify AI and propose guidance that facilitates the integration of AI-tools into the daily operations. Research questions focus on the reliability of existing AI techniques and their implementation and the problem of users trust. It covers the latest trends in the AI domain like responsible AI and explainable AI, their ability to open "black boxes." This thesis contributes to the better comprehension of AI and computational reliability for non-specialists. It proposes a systematic approach to the integration AI-tools and draws companies attention to the significant challenges for AI adoption - the lack of users knowledge and necessity to build users trust. These aspects prevent the exploitation of AI-tools and could lead to missing business opportunities. Business leaders and entrepreneurs must take steps toward building AI-friendly culture, and include the latter problems into a digital transformation strategy which should be developed upfront tech initiatives and has to be seen interconnected with the business strategy.