Thousands of years ago, Mencius debated with The Confessor, who said, "Food, sex, and sex." Ren, inner also, non-external also. Righteousness, outer also. Non-internal also. Although this phrase "food color is also" is often mistaken for Mencius's words, it is not only the son who puts color and food in the same important position, even Confucius once said: "Eat men and women, people's great desires are preserved." ”
The importance of the "color" thing is evident.
And this comes from the most primitive and authentic needs of human beings, which has also spawned and promoted the development of the adult industry for thousands of years.
A few days ago, the TRE Taipei International Adult Exhibition was held in full swing, and the participation of the famous Japanese actress Seiko Takahashi became the biggest gimmick. In a few days, the 2018 Asia Adult Expo will also kick off.
According to the "2018Q1 China Sex Toys Special Research Report" released by Ai Media Consulting on May 18 this year, the Chinese sex toys market is in a period of rapid development at this stage, and it is expected that the scale of China's sex toys market will exceed 130 billion yuan in 2020. Previously, Ma Jiajia, a post-90s girl, has attracted much attention because of the entrepreneurship around sex toys, and in recent years, the industry "old drivers" represented by NetEase Spring Wind have also entered the game.
However, there are adult industries that are reasonably and legally in the "bright place" regarding "sex", and there are also obscene pornography that belongs to the illegal and criminal secrets.
The so-called "spring wind" does not close the jade door, and the awareness of safety should be increased.
The position of "security check" in the adult industry was thus created.
<h1>Color with ring</h1>
Figure | Source: Southern Metropolis Daily
Wu Xia sat in front of the computer, quickly sliding the mouse in her hand, and the computer was a number of open windows. She skillfully clicks on the operation review system, and each user's micro-dynamics, activities, photo album pictures and other information will be displayed on the computer. Her job is to select the offending content. On the other side of the Atlantic, Sarah and Wu Xia do similar work, except that she watches more than 8,000 pieces of racist speech and other unbearable photos and videos every day.
This kind of security inspection work is called a yellow detector.
The daily work of professional yellow experts mainly includes viewing various texts, images, and videos to see whether they contain pornography, bloodshed, violence and other content.
The huge workload, the long time to watch uncomfortable content, over time, as a yellow appraiser, not only will their eyesight decline, but also mentally and psychologically affected.
The British media had previously reported that henry and Greg, two employees of Microsoft's cybersecurity team, had taken Microsoft to court. The reason for the prosecution was that they had to view thousands of videos and pictures of violence at work, which led to symptoms such as schizophrenia and depression in the two.
The duo's lawyer said, "Many people can't imagine how many terrible and cruel things Henry has to see in a day." Not only images of children being sexually assaulted, but also murders against children and unspeakable atrocities. ”
In the long run, Henry had sleep disorders, and the pictures he saw at work often appeared in his mind, and his condition became more and more serious, and even began to have symptoms such as hallucinations, schizophrenia and depression.
And that's not an isolated case. Previously, in the United States, the United Kingdom, China, etc., there have been many cases of Huang Shi suing the "old owner", suing them for the serious psychological impact of the position of Huang Shi.
However, this high-risk profession has been labeled as "high salary" and "mysterious" in China, which has aroused the envy of many people for the profession of yellow connoisseur.
Does the yellow master really enjoy the treatment of "annual salary of 200,000"?
It is reported that Facebook's yellow detector salary ranges from $13 to $28 per hour; some domestic network companies, the monthly salary of the position is only 2,000 yuan -4,000 yuan. Moreover, many yellow experts quit after working for a few months to a year.
The high mobility of personnel has also become a significant feature of this position.
<h1>A strongly needed yellow connoisseur</h1>
On the one hand, the interception of pornographers, on the other hand, with the continuous emergence of Internet social media, live broadcasts, short videos, pornography, violence and other content on the edge of the law crazy temptation.
In the live broadcasting industry, there were "live broadcasts to create people" and "yellow eel masturbation" that were angrily criticized by the media; later, there were beautiful photos of minors naked live broadcast exposed by the "Legal Evening News". In the short video industry, there are Kuaishou who were named by the People's Daily for their vulgar content and ordered to rectify; there are also Douyin, which has just launched an international version and gone overseas, and has been banned by the Indonesian government because of the inappropriate content presented, and the overseas road has died.
These social media platforms that rely on content, while their influence, business, and commercialization continue to rise, content has become their most solid armor, but it has also become the weakest weakness. It is not difficult to see through the "cool" Wuwukai and Chen Yifa incidents that in the face of content security, no matter how high the commercial value is, it cannot be tolerated.
Therefore, the yellow detector responsible for content review has become a vital line of defense to ensure content security, and the yellow detector has naturally become the object of expansion of major enterprises.
On April 7, 2018, after being ordered to rectify, Kuaishou released recruitment information to the whole country, recruited 3,000 content review editors, and expanded the review team of 2,000 people to 5,000 people. Overseas, as of December 2017, Facebook's content review personnel have increased from 4,500 to 7,500, and it is planned that by the end of 2018, the number of people in this position will reach 20,000, and Google, the parent company of youTube, a well-known overseas platform, has also expanded its content review team to more than 10,000 people.
The demand for yellow experts in enterprises is urgent and extensive, but the lawsuits and personnel mobility caused by the psychological harm brought by this profession to practitioners have created a situation where it is difficult to recruit people and the cost is high. With the maturity of artificial intelligence technology and the commercialization of various fields such as finance, medical treatment, services, and entertainment, some technology companies see potential opportunities for artificial intelligence in the application scenario of pornography and content review. Alibaba, Tencent, NetEase, Tupu Technology, etc. are one of the representatives.
<h1>AI to identify the yellow, to understand</h1>
In October 2015, Alibaba announced the official launch of "Green Net", another name of the system is "Yellow Map Scorer", that is, the identification of yellow pictures through AI technology.
In the same year, Tencent launched the Vientiane Youtu Intelligent Yellow Identification Service.
NetEase, which is jokingly called "pig farm", is not far behind, and has launched NetEase Easy Shield to provide services such as intelligent yellow identification and violent terror identification.
Founded in April 2014, Tupu Technology has reached cooperation for many first-line Internet companies such as Xunlei, Kugou, I Tu.com, Qiniu Cloud, Petal Network, etc., to customize personalized picture/video recognition engines for them.
Figure | Source: Weibo
So, how does AI act as a yellow detector?
It is reported that AI forensics is similar to many artificial intelligence systems, using a large number of pornographic pictures as materials, training through deep learning algorithms, identifying pornographic information, and establishing a model through samples, and finally the machine uses this as a basis to determine which pornographic pictures are.
The difference between pictures and live video is that the pictures are static, while the video and live broadcasts are dynamic, but there is no essential difference in their identification methods. When identifying video and live broadcasting, the corresponding video frames are extracted from the video content for image recognition by means of interval screenshots, key frame screenshots, etc.
At present, Facebook, Kuaishou, Xunlei, Baidu, etc. have all identified and filtered illegal content such as pornography and violence through THE AI system. Zuckerberg said Facebook's AI system has been able to filter out up to 99 percent of the bad content, but the rest still needs to be processed by the detector. According to the "2017 Annual Information Security Comprehensive Governance Report" released by Baidu, AI systems can achieve more than 95% of information audits.
Although efficiency has been greatly improved, the current AI cannot completely replace the position of the yellow detector.
Yang Zhiming, CEO of Deep Thinking Artificial Intelligence, who is good at deep semantic understanding in multimodality and combining contextual multi-round semantic understanding, explains: "Now, in image recognition or video, for certain specific targets, such as dogs, positions, and people, these recognitions have been very accurate. However, the understanding of image or video scenes, such as puppies lying under a grape rack to cool off, makes it relatively difficult to understand the whole scene. Pornography, violence, blood, etc. are often exactly what need AI to understand the scene. ”
In addition, Yang Zhiming also said that at the technical level, what people usually say about pornography, violence, etc. is itself an abstract and complex concept. Content recognition can only be made if it is technically defined first. However, technicians can use knowledge graphs to describe these concepts, such as violence, and may include tools of violence, victims, perpetrators, etc. After the definition is clear, and then use the scene to understand, the accuracy of the recognition will be greatly improved.
Care for the physical and mental health of the yellow experts, starting from AI.
<h1>end:</h1>
From the Spring Palace, brothels, and golden plums to today's erotic films and adult industries, color and food have always been the fundamental needs of human beings. With the development of the times, the intervention of science and technology in color and sex will not only stay at the level of AI replacing the yellow detector. At present, many technology companies are developing sex robots, virtual lovers, through new materials, AI, VR and other technical means, multi-level and dimensional to meet human needs for sex.
In the future, who can guarantee that human beings will not fall in love with a voice assistant like in the movie "Her"?
Part of the article source: 1. Southern Metropolis Daily "Female Inspector Huangshi: Look at hundreds of pictures in one minute, the visual impact can't be carried for a long time" 2. SuperD Super Multi-dimensional Technology "How does an AI yellow appraiser work?" Computational Vision Technology and Web Video Auditing.
For more high-quality content, please continue to pay attention to magnesium ke network ~~