The aim of the project is to develop a technology platform that uses artificial intelligence to identify targeted disinformation ("fake news"). To this end, "news-polygraph" combines various software tools for the detection of manipulations of images, sound, videos and texts. The result is a networked range of digital tools for use against fake information or information taken out of context. Over the next three years, a total of ten partners from Berlin, Brandenburg and Thuringia will collaborate on research and development, including institutions from science, media, technology and business.
An AI lie detector for media content: What is fact and what is "fake"?
The planned digital "news-polygraph" tools support the media industry by making it easier and faster to check multimedia content for manipulation with the help of AI. In social media, deliberately placed false information has long been a well-known and widespread problem. But not only is the number of manipulated media increasing rapidly, the quality of the fake content is also rising. So-called deepfakes, image, audio or video files manipulated with AI, increasingly appear deceptively real, so that fakes can only be identified with great effort. "news-polygraph" is intended to support media professionals and other affected industries in the future reliably, time-savingly and intuitively in the detection of false information. For reliable detection of manipulated media content, the AI models must be "trained" by the experts for disinformation. How well the training works is being tested in an initial pilot project in the journalistic field. Other areas of application are diverse and range from PR agencies to public institutions, insurance companies or investigative authorities.
A strong alliance with partners from science, media, technology and business
Ten partners from the region have joined forces for "news-polygraph" and form an interdisciplinary research alliance. Through the three scientific partners - the German Research Center for Artificial Intelligence, the Fraunhofer Institute for Digital Media Technology IDMT and the Technical University of Berlin - "news-polygraph" is closely linked to basic and applied research. The entrepreneurial orientation of the project is strengthened by the five corporate partners Crowdee, delphai, neurocat, transfermedia production services and Ubermetrics Technologies. The concrete pilot project of "news-polygraph" will be implemented by the two user partners Deutsche Welle and Rundfunk Berlin-Brandenburg, who will test the developed prototype with their editorial fact checking experts. "The fight against disinformation is the greatest journalistic and social challenge of our time. The fast and reliable verification of media content is therefore becoming increasingly important in news production, but also in other fields of application. With 'news-polygraph', we want to develop a broadly applicable, intuitive solution for this purpose," says Wilfried Runde, spokesman for "news-polygraph" and Head of Research and Cooperation Projects at Deutsche Welle. News-polygraph is expected to officially enter the implementation phase in the spring of 2023.
The news-polygraph research alliance consists of ten partners, including five companies Crowdee GmbH, delphai by AtomLeap GmbH, neurocat GmbH, transfermedia production services GmbH and Ubermetrics Technologies GmbH, three science partners German Research Center for Artificial Intelligence GmbH, Fraunhofer Institute for Digital Media Technology IDMT and Technical University of Berlin, and two user partners Deutsche Welle and Rundfunk Berlin-Brandenburg.