Amazon and Twitter take steps against deepfakes
The need to consider synthetic media and its dissemination over social media has been acknowledged by Twitter, which plans to enter the fray against deepfakes with “a new policy to address synthetic and manipulated media,” after consulting with experts and the public.
Euronews reports Sam Gregory, programme director for video journalism advocacy group Witness, lauded the initiative but suggested an expanded scope.
“Deepfakes are mobilised to get journalists to share false information and to deceive them … so having journalists who use Twitter respond is a good thing, but it is not enough,” he said. “I think it’s critical that these discussions are global and include those people who have already been harmed.”
Gregory also told Euronews that Twitter needs to be careful in explaining the problem to the public to avoid causing public panic.
Amazon Web Services (AWS) meanwhile has announced it is joining a collaboration between Facebook, Microsoft, academics and other experts to encourage innovation in deepfake detection.
AWS, meanwhile, will contribute up to $1 million in credits for its services to the Deepfake Detection Challenge (DFDC) it is now a partner in, according to a company blogpost.
“Building deepfake detectors will require novel algorithms which can process this vast library of data (more than 4 petabytes),” write AWS Machine Learning Solutions Lab Vice President Michelle Lee in the post.
AWS will also provide access to the the Amazon ML Solutions Labs to experts and solutions architects participating in the challenge for technical support and guidance. All participants can request minimum credits of $1,000 on AWS, and additional grants of up to $10,000 will be awarded to promising entries.
Amazon has recently joined a chorus of voices calling for U.S. government regulation of facial recognition and new data protection legislation.
The DFDC is being launched at the NeurIPS Conference in December.
algorithms | biometrics | deepfakes | facial recognition | fraud prevention | social media | spoof detection | voice recognition