03619ntm a22005417i 4500 000703479 CZ-PrVSE 20240127174846.0 m d cr n|||||||||| 240127s2023 xr fsbm 000 0 eng d NEZPRACOVANÝ IMPORT ABA006 cze ABA006 ABA006 rda Kuska, Petr ISIS:135776 dis Úvod do generativní AI : Pohled na syntetické tváře eng Unmasking generative AI : A look at synthetic faces Petr Kuska 2023 ?? stran : digital, PDF soubor Vedoucí práce: David Chudán Diplomová práce (Ing.)—Vysoká škola ekonomická v Praze. Fakulta informatiky a statistiky, 2024 Obsahuje bibliografii Textový (vysokoškolská kvalifikační práce) Rok obhajoby 2024 This thesis describes the foundations of generative artificial intelligence with a detailed explanation of three generative models – variational autoencoders, generative adversarial networks and diffusion models – and their applications. Furthermore, it discusses the risks and challenges of generative models and the proposed regulation of AI worldwide. It also investigates whether people can identify AI-generated images of faces. An online survey (N = 333) was conducted to test people's ability to distinguish between real and fake faces. The average accuracy of participants was 59.9% (95% confidence interval [58.1%, 61.8%]), which is above chance performance of 50% and is in line with previous research. The analysis showed that having experience with generative AI tools improved the detection ability with a statistically significant difference (p < 0.01) in the means of the inexperienced and experienced samples. The analysis showed no significant difference in accuracy across various screen sizes on which the participants viewed the faces. Seven online tools for detecting AI-generated images were tested, identifying one high-performing tool. In another experiment, three automated machine learning models were trained to classify real and AI-generated faces, which showed that although progress has been made in AutoML image classification, the models were not robust enough and underperformed the best online tool. The survey data and more can be found at https://github.com/mrpekus/master-thesis. Způsob přístupu: Internet znalostní a webové technologie [obor dipl. práce] diplomové práce fd132022 czenas master's theses eczenas diffusion model automated machine learning generative adversarial network generative artificial intelligence real vs fake face variational autoencoders AI-generated image detection Chudán, David js20100609002 ths Zamazal, Ondřej ISIS:7282 opn Vysoká škola ekonomická v Praze. Fakulta informatiky a statistiky kn20010709399 dgg https://insis.vse.cz/zp/82627/podrobnosti VŠKP v InSIS https://insis.vse.cz/zp/82627 Hlavní práce https://insis.vse.cz/zp/82627/posudek/vedouci Hodnocení vedoucího https://insis.vse.cz/zp/82627/posudek/oponent/80648 Oponentura https://insis.vse.cz/zp/82627/podrobnosti dc:identifier NEPOSILAT VSKP vse82627 240123 82627