Deepfakes, a form of synthetic media is the artificial production, manipulation, or modification of images, video, and audio using artificial intelligence. Nowadays the technology is used to create fake persona’s with malicious intentions to deceive the public.
Deep Learning (ML) Software
Deep learning software a form of artificial intelligence is often used to replicate or modify content.
It is a form of Machine Learning technology based on artificial neural networks with representation learning. The term “deep” in deep learning refers to the use of multiple layers in the network.
The neural networks automatically “learn” the characteristics of anyone’s face (or voice) from hundreds or even thousands of examples. Once the characteristics are learned, the AI applies the transformations automatically, creating an almost exact replica.
Although it is often associated with malpractice, Deepfake technology also has its positive uses.
Positive Uses of Deepfake Technology
The underlying technology has been used to add visual effects to Hollywood movies, creating lifelike human avatars for customer services, generating fun social media clips, and more.
On the contrary, it has also been used to create political misinformation, fake celebrities, and to commit fraud.
The technology behind the creation of deepfakes is a combination of new AI-based manipulation tools, traditional CGI and special effects. In reality, technologies behind synthetic media have been in development since the late 1990s.
Historical Overview of the Origins of Synthesized Media
In 1997, Video Rewrite synthesized new lip movements from a separate audio track to make actors appear to mouth words they didn’t actually say. This marked the first significant use of artificial intelligence (AI) to manipulate video in a convincing way.
Today, software companies offering solutions to synthetically generate video and audio is readily available. Shamrock and Ctrl Shift Face are YouTube channels dedicated to Deepfakes that have emerged with large followings. There are also easy-to-use deepfake apps which are commonly used to create entirely synthetic celebrity profiles on Instagram.
How to Detect Deepfake Content
According to research conducted by Oxford Academic, three sources of information can used to accurately detect deepfakes: context, audiovisual imperfections (i.e. technological glitches), and content.
An example is a popular deepfake video featuring former US President Barack Obama.
Technical imperfections are usually a detection of that the content has been manipulated in one way or another.
Other indicators to look out for are unnatural lip movements, asynchronicities of audio and lip movements, or traces of digital rendering.