AI is coming for voice actors’ very talent

AI developers and users continue to challenge the idea of ownership.

In one instance, researchers have found that could get algorithms to create photo-realistic images of real people and recreate copyrighted work.

According to reporting by the technology publication MIT Technology Review, Stable Diffusion and Imagen worked from captions created by researchers from Princeton, ETH Zürich, University of California – Berkeley, DeepMind and Google.

They compared the generated images against reproductions of original images in a model’s training set. In all, they found 100 copied images. Among other things, that finding means AI can, indeed, violate copyrighted work.

According to the Technology Review, the models can “memorize images in their training sets.”

In a second example, voice actors reportedly are feeling pressure give up the rights to their own biometric and tool.

Companies want to feed their voices to AI in a way that would make them unnecessary, according to reporting by technology trade Vice.

To date, much of the focus has been on bad actors approximating people’s voices – usually the voices of very famous people – saying or reading material that they never said. Context is key here, as in the case of actress Emma Watson reading Mein Kampf, according to technology trade publication Motherboard, which shares a platform with Vice.

Voice actors are a noticeable and growing segment of entertainment: animated TV shows and movies, advertisements, games, etc.

The president of the U.S. National Association of Voice Actors has told Motherboard that clauses taking possession of a person’s replicated voice are prevalent.

Article Topics

AI  |  biometrics  |  research and development  |  synthetic voice  |  voice biometrics  |  voice recognition