A group of Russian scientists developed a ‘deep fake’ algorithm that reportedly only needs a handful of photos of a person to create a realistic ‘video avatar’ that will say whatever a user wants. The coders claim they seek to improve the virtual-reality environment.
An article authored by a group of students and post-docs from Russia’s Skoltech has set a citation record on Altmetric Top 100 – an online service that tracks online activity regarding scientific publications in the social media.
The article, which focuses on a system that would allow for the synthesising of realistic-sounding but completely fake speech while mimicking anyone – known online as ‘deep fakes’ – became the most quoted article in the entire 7 years of the website’s existence, the institute said.
Dmitry Lempitisky, a lead author in the study, added that thanks to a deep-learning process based on multiple celebrity photos, the new algorithm needs only a small number of photos of one person to generate a ‘realistic video avatar.’
The institute commented said it was proud of its young researchers, who have sparked an unprecedented wave of interest by their work.
This year, Altmetric Top 100 included studies published in 43 scientific journals, government websites and other sources. Currently, Harvard University and the journal Nature lead the ratings with 11 and 12 articles, correspondingly.
The website tracks online reaction to research posted online by analysing discussions at scientific Internet forums, social media and blogs, as well as in the media and in government documents.
Sourse: sputniknews.com
0.00 (0%) 0 votes
