Maybe you still remember that in June last year, a group of Chinese people used special effects to change their faces and began pretending to be Russians to sell goods.
From indoors to outdoors, from cities to rural areas, they are active in every land where they can pick up their mobile phones to shoot videos and broadcast live broadcasts. They speak Mandarin with a rich dialect and post links in product windows.
Advertisement
There is nothing ruthless in their videos and live broadcasts.
To sum it up in one sentence, “I, a Russian, like China and make money.”
I thought that with the end of that report, the story of Chinese fake Maozi came to an end.
But I never expected that in 2024, this routine would undergo a 360-degree upgrade.
Advertisement
What exactly happened?
In the past two days, a Ukrainian girl named Olga Loiek posted a video, saying that a friend sent her a link. When she saw it, she was stunned.
Hey guys, why did I get the goods on Kakadai in China? ?
It turns out that there is such an account on Xiaohongshu. Tiantian is a blond beauty who looks exactly the same. She talks about the everlasting friendship between China and Russia in a different way. She can't do without thanking China in three sentences. At the end of the video, I will also introduce you to Russian food.
And the old guys who watch the video naturally have to take care of the big girl's business.
Yes, the talking “face” seen by millions of Chinese viewers is Olga Loiek.
On the other hand, Olga has not made even half a dime on YouTube.
What’s even more embarrassing is that, not to mention having his face stolen by others and using it to carry goods, as a Ukrainian, his persona is still that of a Russian, and the goods he carries are all Russian specialties. . . Who can I ask to explain this?
If you can't empathize with this girl, you might as well imagine that an American is standing in your face, pretending to be Japanese, and pretending to be Japanese in a different way.
Then he also used your face to make money from other Americans that you may not be able to make in your life. In the video comment area, the audience wrote “Hello Japan, love comes from the United States.” .
Are you angry?
Olga Loiek is not the only unlucky person. The face of another internet celebrity from Sweden, Rana Blackley, was also transplanted onto a Chinese face.
In the video, “she” talked about the hard work and kindness of the Chinese people, and how she would marry a Chinese husband and live in China for the rest of her life.
After watching the video, netizens didn’t know that they were about to be raped, and asked her which city in China she liked.
After randomly checking several accounts displayed by Olga Loiek, I found that some of these Chinese accounts pretending to be foreigners have been canceled on major platforms, but some are still at large.
But there is only a thousand days to be a thief, and a thousand days to prevent a thief.
The development speed of AI technology is obviously faster than the current response speed of human beings.
A year ago, AI technology was still the object of ridicule by netizens. Netizens would only feel that the uncanny valley effect in this life has been exhausted.
Now, however, AI technology can produce effects that are as real as fake.
Changes always happen quietly. Today, with the advancement of AI technology, eyes and ears are gradually losing their ability to distinguish.
Human perception is being “degraded” under the impact of the information age.
Seeing this, I guess many missionaries will feel that these things are too far away from them. For example, the fake Ukrainian girl, a knowledgeable person will know it at a glance, and she can't deceive herself at all?
If you pay more attention to game chat, then you may still remember a video blogger named Director Ao. He basically lost his reputation due to several waves of public opinion.
Last year, he stopped updating Bilibili and Douyin, and quietly disappeared from netizens' attention lists.
But just two days ago, a station B UP boss used self-developed AI text-to-speech technology, combined with his original copywriting and editing style, to imitate the entire UP boss. Director Ao came out and showed it to me. In a daze.
“Necromancer” and “Successfully summoned souls” are all posted in the barrage.
In the URL posted on the UP homepage, you can also find more weird AI voices, such as Ding Zhen, Kobe Bryant, Sun Xiaochuan, and some celebrities who know that this website will probably issue a lawyer's letter.
So, how many real people’s voices do you need to collect to use AI to generate a person’s voice?
In the video introduction, the UP master gave the answer: 5 seconds.
That's right, you only have to speak for 5 seconds and your voice may have been cloned.
If the above technologies can be used to develop productivity and promote social progress, then it can barely be considered the best outcome.
But you have to know that the current AI technology has become increasingly sinking.
On the first day, a powerful AI software was released, and on the second day, a blogger gave you a tutorial on Douyin. It was a step-by-step tutorial, and everything was clearly arranged for you from getting started to mastering it.
It is undeniable that this is a good thing, but it also amplifies the dark side of people's hearts little by little, and finally forms a public incident.
A few days ago, AI-generated fake indecent photos of Taylor Swift went viral across the Internet and even alarmed the White House.
White House press secretary Karina Jean-Pierre warned that the spread of AI-generated photos was “alarming” and urged social media companies to prevent the spread of such misinformation.
Why is Taylor so explosive? After all, as early as the PS era, many women were suffering from face-changing problems, right?
But in the AI era, all details in photos can be carefully customized.
This time, Taylor was photographed in the rugby locker room just because she had a rugby star boyfriend. . .
Who else is not impatient?
Later, due to too many similar pictures, Twitter temporarily blocked related searches for Swift.
Why is the whole United States becoming nervous because of this matter? Because everyone knows that today it is Taylor, and tomorrow it may be him.
Moreover, the bad things AI does are not limited to the third level. Its power is greater than you think.
In Hong Kong, China, an AI “deep forgery” fraud case occurred.
The scammers used AI technology to replace the voices and faces of senior executives from the headquarters of multinational companies collected online with their own, and then simply defrauded the company's Hong Kong branch of HKD 200 million, and even the company's financial records. There was nothing unusual about the boss's face or voice.
Do you think AI can only reach this level?
Don’t forget that in 2022, a video of Ukrainian President Zelensky “calling on Ukrainian soldiers to lay down their weapons and surrender” was circulated online.
This very cunning video forced the Ukrainian Ministry of Defense to stand up and declare that “this is impossible.”
It can be seen that from individuals to groups, and even wars between countries, every loophole that can be exploited has become a breakthrough for AI.
What’s even more frightening is that so far, there is no particularly good way to prevent residents from using AI to do bad things.
An insider in the network security industry, i Chunqiu’s friend Xiaoao (pseudonym), told Shichao that there are two main difficulties in fighting AI crimes, one is the technical level, and the other is the economic level.
On a technical level, it is difficult to identify content generated by AI. Especially as AI develops, the probability of being able to identify it with the naked eye becomes smaller. In this case, you can only use a machine.
From an economic perspective, it is not difficult to use AI to commit crimes and it does not cost much. However, the cost of using AI to counter AI is high. You have to spend money and effort on research.
At this stage, many regulatory agencies and companies at home and abroad are trying to find ways to tether AI. Last year, the Ministry of Public Security found a gang that used AI to write ransomware viruses.
There are many tools on the market for detecting AI-generated content, but the effectiveness is not yet clear.
For example, OpenAI previously launched a text detection tool, but it was secretly removed from the market because the detection error rate was too high. . .
Two years ago, several technology companies such as Microsoft and the BBC worked together with the media to establish the Content Source and Authenticity Alliance (C2PA).
With the C2PA agreement, who created the content and historical modification records will be recorded. It is equivalent to putting an invisible watermark on the generated content.
But in Xiaoao’s words, C2PA is a thing that protects against gentlemen and not against villains. Big companies abide by the rules. This does not mean that many small models created by open source will also abide by the agreement.
Moreover, C2PA still has many bugs. For example, the watermark disappears as soon as you take a screenshot. For it to be effective, several platforms must use the protocol.
So a very cruel reality is that,At this stage, we cannot prevent AI from doing bad things.
The key is that AI technology is developing too fast. The content generated by AI is becoming more and more realistic, but the threshold for generation is getting lower and lower. Just like after Sora came out, everyone was saying “reality no longer exists.”
It is undeniable that subsequent legislation and anti-AI technology will definitely keep up, but in this situation where the enemy is in the dark and we are in the open, supervision will be very passive.
This is like a cat-and-mouse game. AI knows the rules and will go around them. Humans then set new rules. This game may never end.
All we can do, apart from being more discerning, is to download the National Anti-Fraud Center App.