Google steps up battle on ‘deepfakes’ with data release

SAN FRANCISCO: Google is stepping up its efforts to fight "deep fake" on Wednesday by releasing new data to help detect artificially crafted video.

Technology giants are releasing "large datasets of visual deep fakes that we have made" that can be used as a benchmark to determine whether video has changed.

This is happening, with growing concern that deep fakes can be used to discord or confuse election campaigns.

Some analysts expect realistic deep fakes to be used to disrupt the 2020 US elections.

"Last year we created hundreds of videos with paid and consent actors to create this dataset," said Nick Dufour of Google Research, a separate research unit of the Google Parent Alphabet, and Andrew Gully of Jigsaw.

"We created thousands of deep fakes from this video using the publicly available deep fake generation method. Real and fake result video is a contribution we made to directly support deep fake detection."

Earlier this year, Google released synthetic audio data in the same effort and worked with industry partners to detect and prevent deep fakes.

"Deep fake detection is one of the most important challenges before us," said Sundar Pichai, Google's chief executive officer.

Yasmin Green, the director of jigsaw research, told the researchers that it meant "the largest and most diverse public data set."