With advancements in artificial intelligence reaching unprecedented heights, a new manipulative tool has been created. One capable of seamlessly swapping faces and voices to an extent that distinguishing genuine scenarios from fake ones is a challenging task. This tool, Deepfake, stands as a dangerous threat, particularly in the context of security.
What are deepfakes?
Deepfake refers to artificially created videos made by advanced artificial intelligence and machine learning techniques. It can seamlessly replace faces in a video with those of others. However, it goes beyond mere facial replacement. Nuanced modifications of facial expressions and realistic mouth movements are also a part of the creation of Deepfakes. This sophisticated technology also allows the integration of audio tracks, enabling the manipulated individuals in the video to convey words different from the original recording. So, deepfake is generally a term for realistic editing or re-creation of video that allows you to change or emulate facial expressions in a very realistic way.
A deepfake may portray someone as saying or doing something they never actually said or did. Deepfakes rely on artificial neural networks to recognize patterns in data. Producing a fake photo or video usually involves inserting hundreds or thousands of images into an artificial neural network and training it to identify and reconstruct patterns (usually faces). Techniques behind the creation of Deepfake are many.
Among the current techniques we’re using we can mention, for example, computer lip synchronization, manipulation of attributes, exchange or manipulation of faces (substitution of expressions and features), generation of completely new fictitious faces, voice synthesis in the context of creating audiovisual fakes, synthesis of the whole face, etc. The different methods of manipulation categorize depending on the technical and technical complexity.
Deepfakes as political game changer
The inseparable link between Deepfake technology and security threats has led us in an era where misinformation, propaganda, and hybrid actions pose significant challenges to the fabric of democratic discourse. In the context of hybrid warfare, propaganda is a systematic, targeted activity aimed at influencing the broadest masses (the target audience) with certain ideas and courses of action, where such winning of a broad audience is using to support the policies of a particular government.
Propaganda is then very often based on a form of demagogic advocacy and recommendation of ideas and opinions with the aim of gaining a public following for effective political struggle, which may also escalate to warfare. The potential of Deepfake to influence public opinion has elevated from a mere technological innovation to a potentially powerful force. Deepfake has transformed what might have been perceived as a superhero into an unsettling villain in the realm of politics. The ease with which realistic content can be spread, often indistinguishable from authentic information, raises urgent concerns about the integrity of democratic processes and the very nature of truth in political discourse.
deepfake technology offers innovative possibilities for communication and expression. However, its misuse carries the potential to break trust in institutions, manipulate electoral outcomes, and undermine the core tenets of democratic governance. A crucial task lies ahead: safeguarding the integrity of democratic systems against the looming shadow of Deepfake’s potentially malicious influence.
When an experiment goes wrong
Speech AI startup ElevenLabs opened a beta version of its Voice Lab platform. The version allows users to create entirely new synthetic voices (for text-to-speech) or very faithfully “clone” an existing voice. Voice Lab uses deep learning methods to duplicate and synthesize human speech. This technology can capture a human voice’s subtleties, cadence, and intonation with astounding precision thanks to cutting-edge neural network models. It offers a variety of possibilities, whether you want to use it for independent projects, content development, or even accessibility improvements. However, the risk of misuse is extremely high with such tools and Voice Lab is a great example. In this case it happened unexpectedly soon, just a few days after the beta version was released. [1]
As the capabilities of platforms like Voice Lab continue to evolve, the safeguarding against any misuse becomes very important. The tech community and policymakers proactively address the ethical considerations surrounding the rapid advancement of AI technologies. In doing so, we can collectively ensure that the future of Speech AI is not only marked by unprecedented possibilities but also by a responsible and ethical approach. It could preserve the integrity of the technology and its impact on society.
Resources
[1] A new AI voice tool is already being abused to make deepfake celebrity audio clips [online]. [cit. 2023-12-04]. Available from: https://www.engadget.com
cover image created using Bing AI image creator with DALL-E 3 technology