Reuters studies real-time video newscasts whose conductors are not existing, but created by a deepfake-style technology, that is, that of fake videos but used for deceptive purposes. For the project, the news agency is collaborating with a start-up that deals with artificial intelligence and is called Synthesia.
The project, for the moment in a test phase, takes the data of the scores of the football matches in real time and generates news complete with images. Synthesia and Reuters use a neural network similar to that of Deepfake and based on pre-recorded footage, they transform the text into a video with a non-real conductor.
"Reuters has long been at the forefront of exploring the potential of new technologies to deliver news and information. This prototype is helping us understand how Artificial Intelligence and the media can combine for a new type of product," he explains to the site. TheNextWeb Nick Cohen, head of Reuters' products and new services, but the applications of this technology could be enormous and not only cover the world of media but also that of video services of airports or stations that are used to give information or updates.
Reuters is focusing on this type of technology. A few weeks ago, she teamed up with Facebook to create an online course that identifies deepfakes. The course lasts 45 minutes and is designed to give journalists the tools necessary to detect and avoid fake photos, videos and audio. in turn, the social network owned by Mark Zuckerberg, has entered a coalition composed of several companies and universities called Deepfake Detection Challenge and has committed to investing 10 million dollars to create videos with researchers who explain how to locate videos counterfeit.