Abstract
In the research and popular literature, there are many concerns about the deepfakes that may be generated using generative AI and other tools to mislead people and to compromise them in various ways through deception. This work explores the nature of deepfakes, generative AI, and manual and computational and combined means to identify (and neutralize) potential deepfakes, as described in the academic literature. Deepfakes may have positive effects as well to enhance human creativity and to expand the toolkit for digital making. This work also explores how to create a credible (non-malicious) 2D visual deepfake using a popular generative AI tool to understand some affordances and constraints for this particular “use case”. For this basic work, the idea is only to trick the eyes, not the more in-depth detection methods based on computation.
Key Terms in this Chapter
Deepfake: A digital illusion made up of synthetic media created with various technologies (including generative AI, image editing software, animation software, video editing software, and others).
Large Language Model (LLM): A computational model created by large-scale training data (of text, image, video, and multimodal content) using deep learning methods (like artificial neural networks).
Media Forensics: Techniques for validating or invalidating digital media.
Authentication: Validating the genuine origins of digital media; affirming the pristine feature of digital media.
Pixel: Picture element, a small box of illumination on a screen comprised of red, green, and blue (RGB) wavelengths to create a sensation of color.
Artificial Intelligence: Computer systems that emulate human intelligence and capabilities, such as creativity, among others.