AI learned to recognize newfangled fake videos (video)
AI learned to recognize newfangled fake videos (video)

AI learned to recognize newfangled fake videos (video)

ИИ научился распознавать новомодные фейковые видео (видео) Specialists have found a way to protect celebrities and other people from fake videos. State University of New York introduced a neural network that distinguishes diphtheics. The secret is in blinks. At one time, videos with substituted faces flooded the Internet and what at the beginning seemed funny and funny, after a while, it was already worrying. Politicians can be “forced” to say things that they would never have said. To create such a video, it is necessary to process a large number of photos in order to “remember” a typical facial expression, therefore celebrities most often suffer. However, the photo does not take into account the fact of a person blinking. The engineers took advantage of this by creating a program that determines the fake videos with a 95% probability. Unfortunately, this method will not protect for a long time, over time, scammers will find the opportunity to adapt to the standard blinking of a person (on average once every 2-10 seconds, lasting from 1 / 10 to 4/10 seconds).

Add comment