Understanding Deepfakes
Definition: Deepfakes are AI-generated videos or images, using deep learning to manipulate content, enabling anyone to alter speeches, movies, or even dance routines.
Applications: Primarily found in pornographic content, deepfakes are expanding beyond celebrities to potentially fuel revenge porn, while also being used for satire, spoof, and mischief.
Diverse Formats: Deepfake technology extends beyond videos to create fictional photos and even manipulate audio, spawning a range of deceptive practices from fake journalist profiles to voice cloning.
Creation Process: Involving intricate steps, creating a deepfake requires running faces through AI algorithms, employing encoders and decoders, or using generative adversarial networks (Gans) to generate realistic faces.
Creators: Various entities, from researchers to amateurs, visual effects studios, and even governments, engage in making deepfakes, with an increasing array of tools available, including mobile apps like Zao.
Detection Challenges: Detecting deepfakes becomes complex as technology evolves. Blinking patterns were initially a telltale sign, but improvements in algorithms quickly address and overcome such detection methods.
Potential Harms: While deepfakes may not trigger international incidents, they possess the potential to disrupt stock prices, influence voters, and exacerbate religious tensions.
Trust Erosion: Beyond immediate harms, the broader impact lies in eroding trust, making it challenging for individuals to discern real from synthetic content, with implications for legal proceedings and personal security.
AI Solutions: Ironically, AI emerges as a potential solution to counter deepfakes, with ongoing efforts to develop detection systems and explore the use of blockchain for verifying media authenticity.
Varied Impact: Not all deepfakes are malicious; some serve entertainment or practical purposes, such as restoring voices lost to disease or enhancing cultural experiences in museums.
Shallowfakes: A Cruder Variant
Definition: Coined by Sam Gregory, shallowfakes involve simple editing or presenting videos out of context, impacting public perception with crude yet effective manipulations.
Examples: Instances include a doctored video of Nancy Pelosi and altered footage of CNN correspondent Jim Acosta, illustrating the potential impact of manipulated content on public figures.
Political Use: Shallowfakes are wielded in politics, with instances of the UK’s Conservative party doctoring interviews to influence public opinion during elections.
Increasing Mischief: As technology advances, shallowfakes contribute to a synthetic world, emphasizing the persistent challenge posed by deceptive media practices.
Enduring Presence: The influence of deepfakes and shallowfakes continues to grow, necessitating ongoing efforts to understand, detect, and mitigate their impact on society.
