You may not be familiar with the term DeepFake, but you are certainly familiar with the images below.
The image on the left is the recently popular Yang Mi version of Huang Rong, while the image on the right is Nicolas Cage in an interview. Both images tell the same story: AI face swapping, a topic that has sparked widespread discussion.
AI face swapping first occurred with Nicolas Cage, initially as a humorous prank. However, the story took a darker turn. AI face swapping was maliciously used in adult films, and many well-known actresses suffered as a result. The Yang Mi version of Huang Rong can be considered the domestic version of Cage’s story. Once released, it garnered widespread attention in China due to its seamless results without PS traces.
What was the outcome of the Yang Mi version of Huang Rong, the most popular face-swapping story in China? The author of this incident calls himself “Face Swap Brother.” After the incident was reported by the media, Face Swap Brother gave interviews and spoke on Weibo:
He admitted that after the incident was reported, many people wanted to contact him for video face swapping, most of which involved illegal activities. Face Swap Brother stated on Weibo: “The issue of copyright has made me deeply aware of my mistakes and ignorance.”
Recently, DeepNude has become even more infuriating. DeepNude is an application based on deep neural networks that can undress someone with one click. Previously, this application offered both a client version and a web version. You could choose to pay $50 for the paid version or use the free version, with the difference being that the free version would have a large watermark on the images.
This application was taken down just a few hours after its launch. Subsequently, the developers of DeepNude announced on Twitter the permanent closure of the service. They stated that they created this entertainment project months ago and once thought its spread was within a controllable range. In fact, this application was not very well-developed; it only worked with specific photos. They never anticipated it would spread like a virus to the point where it became uncontrollable.
In the discussions sparked by DeepFake, there are two attitudes towards face swapping applications.First, there is an acknowledgment that technology is innocent, and that face swapping can indeed satisfy certain entertainment needs of the public.Secondly, there is a greater concern about the unpredictability of who will use this capability for what purposes.All of these represent potential risks.However, the emergence of DeepNude has led to a one-sided public opinion.Indeed, this software is quite popular among some people.It is said that even though the professional version of this app costs $50, many are still willing to pay for it.
Society’s attitude towards DeepFake is one of skepticism, while the attitude towards DeepNude is one of condemnation. They have raised concerns about the unreasonable use of AI technology. Technology is innocent, but the negative impacts that arise from its unreasonable use cannot be ignored. Professional lawyers have pointed out that AI face swapping raises issues of infringement of portrait rights and reputation rights. Clearly, DeepNude exceeds this point!
In response to these fabricated images and videos, some governments have already enacted specific laws to combat DeepFake technology. In addition to policy measures, various technical methods can help us identify these fake videos and photos. Researchers have developed algorithms specifically for detecting these forgery techniques. Although the latest algorithms have a detection accuracy of up to 97%, the remaining 3% of false identifications can still lead to a large number of photos leaking, resulting in serious consequences.
The value of technology lies in its ability to meet specific needs of the public. An important premise is that this need must be legitimate. Everything has its duality, and AI is no exception. In today’s rapidly developing AI technology, AI’s capabilities surpass those of the average person in many areas. However, the negative impacts brought about by the unreasonable use of AI technology are increasingly concerning. DeepFake and DeepNude are the most typical examples.
AI applications born from base needs are inherently sinful and should be approached with caution. In the era of mobile AI, the threshold for technology applications has become very low.This fundamentally makes it impossible to eliminate the situation of “unreasonable use of technology.” Especially with the emergence of platforms, this uncontrollability is exacerbated.
In the book “Flaws: The Unicorn on the Wind” there is a viewpoint:Platform models have inherent flaws:Quality control.For example, it is difficult for platforms like Taobao to solve the problem of counterfeit goods, and Didi finds it hard to address driver crimes.AI technology is similarly challenged; the larger the company and the more users it has, the harder it is to eliminate unreasonable uses of technology.
Due to the peculiarities of social interactions, social channels are increasingly monopolized by a few giants.Everyone is on the same social platform.Facebook has nearly 2 billion users, while WeChat, the leading social software in China, has around 600 million users.Against this backdrop, information can be transmitted to a massive audience in a way never seen before.AI technology is the same; technologies like DeepFake and DeepNude have lowered the threshold for using this technology so much that anyone only needs a computer and a phone.How can this not be concerning?
A common person is innocent, but possessing a treasure can lead to guilt; technology is innocent, but misuse is culpable, and the focus should be on regulation. If there is no way to control the misuse of technology, shouldn’t we pause for a moment and not immediately promote it on a large scale? Would it not be more appropriate and responsible to wait until regulatory measures are in place before promoting it? Technology enthusiasts can conduct appropriate research out of curiosity and love for technology, but they must be aware of the uncontrollability and potential risks of technology dissemination. Major technology platforms should be even more cautious in the face of such base demands.
Don’t be evil. Also, do not inadvertently become an accomplice to evildoers. Wouldn’t it be better to focus on doing things that are truly valuable to the public and society?
CariGANs: Automatic Generation of Portrait Comics through Image Style Transfer
N2N Learning: Model Auto Compression Based on Reinforcement Learning
From DQN, DPG, Actor-Critic to DDPG
Analysis:The Mathematical Principles Behind GAN
