Stroje a rasismus - Stanovení moderních etických standardů pro umělou inteligenci

This master's thesis aims to achieve three goals. Support discussions for talks about AI regulations with the state of art data, analyze what is the current state of laws in the largest countries of the world, with the hypothesis being that it is critical and help create classification methods for AI-related definitions, as it is lacking. The reader will get the theoretical basis of AI-related topics from the basics to use cases and the most recent important topics, along with an explanation that AI is a positive thing that, however, needs to be regulated. Afterward, racism will be introduced from two points of view. It will start with classic racism between humans, as human-made data with all of its biases can later be used as a source for a model. Emphasis will be made on why people are racist and what are the positive effects of restraining racism. It will be followed by an introduction to why AI can be considered racist and different kinds of bias that can come into play during all phases of its development and operation. At the end of the theoretical part, two real-life cases of AI racism will be introduced to solidify further the need to create rulesets that will address new challenges caused by the technology. In the practical part of the paper, I analyze AI laws and white papers to inspect how much they meet the conditions set up by the theory of this paper. In conclusion, the hypothesis of a troubling state of the matter has not been accepted, albeit room for improvement has been found, and recommendations have been given.