Deepfake Technology, Market Analysis
Emergence of deepfake technologies has profound implications on public, government and media market segments. There are some potentially benign or beneficial uses of this technology.
Deepfakes are powerful tools that can be used for exploitation and disinformation. Deepfakes could influence elections and erode trust but so far have mainly been used for non-consensual pornography. The underlying artificial intelligence (AI) technologies are widely available at low cost, and improvements are making deepfakes harder to detect. A deepfake is a video, photo, or audio recording that seems real but has been manipulated with AI. The underlying technology can replace faces, manipulate facial expressions, synthesize faces, and synthesize speech. Deepfakes can depict someone appearing to say or do something that they in fact never said or did.
Deepfakes rely on artificial neural networks, which are computer systems modeled loosely on the human brain that recognize patterns in data. Developing a deepfake photo or video typically involves feeding hundreds or thousands of images into the artificial neural network, “training” it to identify and reconstruct patterns—usually faces.
Deepfakes use different underlying AI technologies—notably autoencoders and generative adversarial networks (GANs). An autoencoder is an artificial neural network trained to reconstruct input from a simpler representation. A GAN is made up of two competing artificial neural networks, one trying to produce a fake, the other trying to detect it. This competition continues over many cycles, resulting in a more plausible rendering of, for example, faces in video. GANs generally produce more convincing deepfakes but are more difficult to use.
Researchers and internet companies have experimented with several methods to detect deepfakes. These methods typically also use AI to analyze videos for digital artifacts or details that deepfakes fail to imitate realistically, such as blinking or facial tics.
- Entertainment. Voices and likenesses can be used in a movie to achieve a creative effect or maintain a cohesive story when the entertainers themselves are not available.
- E-commerce. Retailers could let customers use their likenesses to virtually try on clothing.
- Communication. Speech synthesis and facial manipulation can make it appear that a person is authentically speaking another language.
- Data needs for detection. Deepfake detection tools must generally be trained with large and diverse data sets to reliably detect deepfakes. Technology companies and researchers have released data sets to help train detection tools, but the current data sets are not sufficient by themselves. Detection tools must be constantly updated with data of increasing sophistication to ensure that they continue to be effective at detecting manipulated media.
- Detection is not yet automated. Current tools cannot perform a complete and automated analysis that reliably detects deepfakes. Research programs are currently working on means to automatically detect deepfakes, provide information on how they were created, and assess the overall integrity of digital content.
- Adaptation to detection. Techniques used to identify deepfakes tend to lead to the development of more sophisticated deepfake techniques. This “cat and mouse” situation means detection tools must be regularly updated to keep pace.
- Detection may not be enough. Even a perfect detection technology may not prevent a fake video from being effective as disinformation, because many viewers may be unaware of deepfakes or may not take the time to check the reliability of the videos they see.
- Inconsistent social media standards. The major social media companies have different standards for moderating deepfakes.
- Legal issues. Proposed laws or regulations addressing deepfake media may raise questions regarding an individual’s freedom of speech and expression and the privacy rights of individuals falsely portrayed in deepfakes. Moreover, potential federal legislation aimed at combating deepfakes could face enforcement challenges.
Interested in more of this research and a chance to learn about opportunities in the emerging deepfake technology market? Our report features data-driven insights from our market intelligence platform.