December 23, 2012 - In recent years, artificial intelligence (AI) The rapid development of technology has brought a lot of convenience to society, but it also comes with security risks. According to IT home understanding, one of the depth of forgery technology has brought no small hidden danger, its ability to generate highly realistic but not real video or image, has led to a series of social problems.
According to CCTV News, this year, there have been numerous cases of fake celebrities using AI deep counterfeiting. For example, not long ago there were unscrupulous people who faked the face and voice of Dr. Zhang Wenhong through AI technology and used it for live streaming to bring goods, selling more than 1,200 pieces of protein bar products.
In addition, celebrity Andy Lau and entrepreneur Lei Jun have also become targets of AI forgery. Andy Lau's voice has been maliciously used to increase traffic to short videos, and his movie company has issued several statements to warn the public.
Experts point out that current AI technology can easily clone someone's face and voice through deep learning, even with just a photo and a voice sample. However, this technology is not flawless. For example, fake videos usually cannot be streamed live in real time, while there are flaws in details such as facial contours and background blending.Viewers can spot the breaks through close observation, especially in terms of lip-synching and voice matching.
Unauthorized use of another person's image or voice is a copyright infringement, according to the Civil Code and the Cybersecurity Law, experts said.Publishers may face penalties including damages and criminal liabilityThe AI is a new technology that has been developed by the University of California, Berkeley. At the same time, consumers who are affected by AI counterfeitCelebrities lead the wayIf you have been cheated, you can claim "one refund and three compensations" according to the Consumer Protection Law.
Experts also called on short-video platforms to take the main responsibility to take down offending content in a timely manner and take credit management and punitive measures against publishers.