Recently, some netizens uploaded a video of Yang Mi "changing his face". In the video, the producer used AI technology to change Yang Mi’s face to Zhu Yin’s character Huang Rong in "The Legend of the Condor Hero". The video immediately triggered a heated discussion on the whole network, and there were continuous hot searches. Users said that the video had no sense of inconsistency, and even "thanks to technology, Yang Mi finally has acting skills". In addition to the large-scale amazement of the Internet, this black technology of "changing faces every minute" also makes people think carefully and feel terrifying.
Yangzi Evening News/Eye Raising Reporter, Kong Xiaoping
[Event playback]
"Yang Mi’s version of Huang Rong" has no flaws
In the 94th edition of "The Legend of the Condor Hero", Zhu Yin played Huang Rong, who was very smart and playful, and many scenes could be called classics. Bilibili, a UP master named "Brother Face Changer", used AI technology to replace Yang Mi’s face with "Huang Rong" in the play. Looking at it, "Yang Mi’s version of Huang Rong"’s expression and body movements were very smooth, and no flaws could be seen at all. Because it was too realistic, it quickly caused an uproar on the Internet. It is reported that the face-changing brother has previously released a number of videos of changing faces for film and television characters, including Dragon Mom in A Song of Ice and Fire, Rose in Titanic, Liu Piao Piao in King of Comedy, Laura in Tomb Raider, etc., and the replacements are all Yang Mi.
This is not an exception
After the video appeared, netizens went from stunned to amazed. There was even a topic on the Internet called "Replace Zhu Yin’s Huang Rong with Yang Mi’s face", and the reading volume actually soared to 110 million, and there were 25,000 discussions. Some netizens said that with the blessing of technology, anyone can become an acting school. Will face-changing technology become a new shortcut for performers? In the future, celebrities don’t need to be present for filming, they can directly sign a license contract "allowing face-changing", and they can produce film and television works.
In fact, "AI face-changing" did not start with @face-changing brother, but attracted wider attention because of "Yang Mi’s version of Huang Rong".
The tool used by @Face Changer to make this realistic video was called deepfake. It appeared abroad in 2017. Some people used celebrities to make some indecent videos. Unfortunately, the celebrities who were recruited included Gail Gadot, Scarlett Johansson, etc. At that time, it caused a lot of controversy, and many platforms explicitly banned the dissemination of such videos.
"It’s simply a technical obscenity, ask for rectification"
After deepfake was introduced into China, some netizens in China have also made face-changing videos. For example, the female live streaming host that otaku men love to see has been replaced by a domestic popular flower, with hundreds of thousands of views. Hollywood actress Scarlett Johansson once predicted that "it is only a matter of time before anyone becomes the target of this technology."
Therefore, the reporter saw that many sober netizens also realized that public figures only bear the brunt of "technology". If someone maliciously uses and abuses, anyone can become anyone, and it is not safe for everyone. Some netizens said that "face changing" is simply a naked "technical obscenity". Countless netizens called on relevant departments to "seek rectification"!
The publisher has removed the relevant videos from the shelves.
On the 27th, the video publisher @facechanger issued an apology on Weibo, saying that it had communicated with Yang Mi’s official studio and removed the relevant video. He also said that ordinary AI face-changing technology may only be used for entertainment purposes, and higher-level AI technology such as military use has high security, and current civilian AI technology will not pose a threat. He also said that this is a warning to everyone that everyone should respect copyright and portrait rights and focus on the technology itself.
[Chain reaction]
Thinking about it carefully, I feel very terrifying. Is it safe to pay by brushing your face?
This incident has undoubtedly caused a certain negative impact on society, but it is terrifying to think about it carefully. Nowadays, AI is used in various scenarios such as autonomous driving, facial recognition, smart cities, and smart homes, including facial recognition systems. However, once the back-end control system is attacked, there will be a large-scale and chain collapse, and the consequences of AI getting out of control will be unimaginable. According to online reports, in January 2019, the Dutch Deeptrace Laboratory released a deepfake development report saying that the number of searches for "deepfake" keywords in 2018 increased by 1,000 times compared with 2017.
On the night of the 27th, @AliSecurity made a popular science and response on Weibo. Jue Ao, an advanced algorithm expert at AliSecurity Turing Lab, pointed out in the video that the current mainstream face-changing technology is based on deep learning models, and its core is an "autoencoder". Specifically, through face detection technology, it takes thousands to tens of thousands of faces to capture the faces of two people in the video, such as Zhu Yin and Yang Mi. After training, enter Zhu Yin’s face to generate Yang Mi’s face.
What kind of problems will the application of this technology cause? Jue Ao said that in addition to the moral and ethical issues caused by face changing, the "black and gray industry" uses AI to change faces, which is a common attack method. "In the facial recognition system, the technology of recognizing face changing is called live detection technology, which is an attack and defense confrontation technology, that is, using AI means to solve AI attacks." Jue Ao also said that through deep learning technology, training models can be identified by the difference in texture, skin color, reflection and other characteristics of the two. It can be achieved in milliseconds to find fake faces, and the current recognition accuracy rate is 99.9%. At the same time, he also pointed out, "In recent years, due to the rapid development of deep learning, the technology of attack and defense has been constantly improving. This is a process of continuous and dynamic confrontation, and we have the confidence and ability to do it well."
To this end, @Ali security said that Ali’s facial recognition live detection technology can be identified in milliseconds, the success rate is as high as 99.9%, and there are still prevention and control technologies.
[Reporter interview]
New technologies cannot break through the legal bottom line
Xu Difeng, a professor and doctoral director at Nanjing University Law School, said in an interview with reporters that there have been infringement acts of imitating the voices of celebrities in the past. He believes that there is no right or wrong for "AI face-changing" as a technology. AI face-changing celebrities are similar to the behavior of using other people’s photos without permission, and they are all suspected of infringing the right of portrait. Platforms and individuals who use technology unreasonably should be bound by the law.
Where are the boundaries of the application of "face changing"? Xu Difeng said that in the past, it was PS photos, but now face changing is PS videos, and the legal awareness that other people’s photos and videos cannot be used casually is also gradually developing. In addition, this AI face changing technology has certain latent risks, and there are currently legal norms applicable to it.
In the interview, some experts in communication also said that the development of AI technology allows ordinary people to enjoy the convenience brought by technology, but new technologies should drive the renewal and iteration of benign cultural creativity, avoid using crooked roads, and do not break through the bottom line to challenge public order and good customs.
