Author | Kazi
Editor | Curly Yu Tian
Header picture | Midjourney
* Today's header picture was created using Midjourney, with the keyword "a girl, sitting in a dark room crying, illustration style"
The distance between us and evil is related to tools.
AI, this (perhaps) the most powerful tool in history, is also fully armed in the hands of evil people. For ordinary people, the addition of AI has made all kinds of scams and harms even more difficult to guard against.
Just imagine, any figure or movement, that face could be your own, is it scary?
Selfies and videos posted by individuals on social platforms can easily be processed by AI face-changing and AI painting technologies, and used by criminals as "material" for indecent pictures and videos, to be sold or used fraudulently.
Think about it again, the information you see overwhelmingly is all produced and distributed for a certain purpose. Is it scary?
It is difficult to tell the authenticity of various screenshots and chat records circulated on the Internet. When AI is added to "creation", the style can be imitated, the logic can be self-consistent, and it can also tirelessly mass produce, information pollution, and achieve it with one click. .
The following evil incidents all have the application of AI.
1. AI-generated soft pornographic photos of children appear on social platforms
On May 16, the Weibo blogger "Cyber Shikuai Appraisal Office" posted that it was discovered that some users on station B had uploaded realistic-style soft pornographic images of children synthesized by AI painting tools. The girls in the generated pictures are scantily clad and behave in an indecent manner, and users will share "picture packages" in comments and updates.
After the incident was exposed, due to reports from users, some videos have been removed from the shelves, and the accounts of the relevant users have been blocked.
Source: Cyber Ten Block Appraisal Office
2. The CP29 Comic Exhibition photo was taken in an indecent manner
On May 6, some netizens contributed to the Weibo blogger "Spring Short Man", saying that a netizen expressed dissatisfaction with the cosplay photos taken at the comic exhibition that day on a forum, so he used AI painting tools and LoRA models (a A fine-tuning model), which "refines" pictures of amateurs into indecent photos of "skirt-up" through partial adjustments.
After the submission was published on Weibo, it was forwarded as high as 21,000 times.
Source: Kasuga Short Man via Weibo
Coincidentally, on May 3, blogger DeadPrince was informed in a private message on Weibo that his selfie was made into a LoRA model by a stranger and uploaded to the painting model sharing website Civil, and someone used the model to draw nude photos.
Source: DeadPrince
On February 16, ChatGPT was discussed in a community owner group in Hangzhou. One owner tried to use ChatGPT to write a fake press release, the general content of which was "Starting from March 1, 2023, Hangzhou will cancel travel restrictions." . Subsequently, this fake news was screenshotted and spread by group members who did not understand the reason, and even became a hot search. In the end, the police intervened in the investigation and the business owner involved publicly apologized.
Source: Zhejiang Voice
On May 13, according to the "Daily Mail" report, a small Spanish film company, Peabody Films, received a call from the famous British actor Benedict Cumberbatch, saying " "Juan Fu" is interested in movie scripts and wants to make a movie together.
However, the film company invited him for interviews several times, but "Juan Fu" refused. He even offered a sky-high interview fee and asked the company to remit 200,000 pounds first.
Feeling something was not right, the film production company confirmed with the "Juan Fu" agency. After confirmation, they discovered that this was an AI scam.
According to the staff of the film company, if they only look at the voice call, since the voice similarity is as high as 100%, they are completely convinced that the other party is "Quan Fu".
On April 26, Internet celebrity "Caro Lailai_" posted on Weibo that she was "face-stealed" by an external social platform, pretending to register her own fake account, and fans can watch her as long as they pay. Indecent video.
She also discovered that many pictures of celebrities, bloggers and even amateurs have been made into indecent videos to make money in this way. They even provide AI customization services. Using the AI technology of online websites, you can achieve "one-stop "Button to take off your clothes".
AI “One-click undressing”
On March 28, “A woman’s subway photo was disseminated by AI with one click of her clothes undressed” became a hot search on Weibo. The woman in the photo was a Xiaohongshu blogger, and the so-called “nude photo” was taken from her in July last year. A photo shared by Yue on the platform.
In the original picture, she was dressed normally, but someone with a bad intention used AI software to "undress" with one click, and deliberately spread it throughout the Internet, causing a bad impact. After the incident broke out, the blogger has sought rights protection from relevant departments.
It is difficult to distinguish the authenticity of AI content, and supervision at home and abroad has been strengthened
Due to the popularization of AI technology, a large amount of AI-generated content has emerged on major social media platforms. However, the real-life pictures and videos generated by AI, and the text manuscripts that are difficult to distinguish between true and false are easily confused if you don’t know much about AI technology. "Being fooled".
For example, some time ago, an overseas netizen posted a set of "Fashion Pope" spoof photos on the Twitter platform and went viral. The "fake pope" in the picture took off his traditional robes and put on a fashionable white down jacket. It is very easy for passers-by who don't know the truth to misunderstand that this is the real pope changing his clothes, but it was actually drawn using the AI drawing tool Midjourney.
Source: @skyferrori via twitter
Because this group of AI-generated pictures were too "hot" to prevent passers-by from misunderstanding, Twitter, as the platform, promptly marked the pictures with "AI-generated".
Looking at China and abroad, many platforms have recently followed up on the supervision of AI content in a timely manner and issued artificial intelligence-related specifications and measures.
1. Douyin releases “Douyin’s Platform Specifications and Industry Initiatives on Artificial Intelligence Generated Content”
Articles 9 to 11 of this industry initiative mention the identification of AI-generated content and the ability to register virtual people, and open user feedback channels.
9. The platform will provide unified artificial intelligence-generated content identification capabilities to help creators mark and facilitate users to distinguish.
10. The platform will provide the ability to register virtual people and protect the registered virtual person images.
11. The platform will provide user feedback channels to facilitate users to report illegally generated content.
Related articles: "Douyin's new rules come into effect!" Virtual anchors are finally "legal", but they must reveal their identities and be driven by real people》
2. Xiaohongshu adds AI creation logo to notes
At the end of April, some Xiaohongshu users discovered that the platform began to mark AI-generated content and asked users to check the authenticity.
Source: Internet
3. Google announced watermarking for AI-generated images at the 2023 Developer Conference
Google announced at the recent 2023 Developers Conference that it will ensure that every AI-generated image of the company has an embedded watermark. Although the watermark cannot be recognized by the naked eye, it can be read and used by software such as Google search engines. The label is displayed to remind the user that the image was generated by AI. This feature will be available in the coming months.
In addition, my country released the "New Generation Artificial Intelligence Ethics Code" in 2021, which involves strategic planning, the formulation and implementation of policies, regulations and technical standards, resource allocation, supervision and review, etc. And in March this year, the Cyberspace Administration of China issued the "Regulations on the Management of Generative Artificial Intelligence Services (Draft for Comments)", which regulates the behavior of generative AI service providers, including the protection of user rights and interests, information security assurance, technology Safety and security, etc.
Looking overseas, the Italian government imposed a temporary ban on ChatGPT in March this year, citing concerns that it violated the European Union’s General Data Protection Regulation. On May 16, OpenAI CEO Sam Altman also testified before the U.S. Senate about the potential dangers of AI technology, and urged lawmakers to implement licensing requirements and other regulations for organizations developing advanced AI systems such as GPT-4.
OpenAI CEO appeared on Capitol Hill in the United States, calling for the regulation of AI. Source: Internet
It can be seen that most of these infringements are still in the information field. On the one hand, from the material point of view, it is relatively ownerless public information. Things that were originally discussed on the streets after dinner, although there are a lot of falsehoods and conjectures, are nothing more than spreading from one street to that alley. But just like social media can spread information As the exile increases, evil people can also use AI to amplify the turbid current; on the other hand, from the perspective of the victims, most of them are ordinary people who have difficulty speaking, and no one will read their statements, and it is even more difficult to send legal letters. Although laws, regulations and platforms are following up on the supervision of AI technology, the cost of rights protection for ordinary people is still very high.
The problem comes back to the bad guys. When they hide behind the Internet, it becomes a must to control the knives in their hands.
ps.Tomorrow will be 520
, as a romantic AI player, how can I miss out on roses?
Reply to the keyword "520" or "521" in the backend of the "Number One AI Player" public account, and you can randomly get a blind box rose generated by AI
The above is the detailed content of Your selfie may be included in the 'indecent image library', and AI's evil is actually very close to you!. For more information, please follow other related articles on the PHP Chinese website!
![[Web front-end] Node.js quick start](https://img.php.cn/upload/course/000/000/067/662b5d34ba7c0227.png)
