"Deepfakes" are created by misusing AI: These cleverly crafted fakes even control human thought

Recently, fake news has spread that actor Kato Seishirō, actress Aoi Yu, and Yamamoto Mizuki made comments in support of "Reiwa Shinsengumi" and Taro Yamamoto, causing a stir with their agencies and political parties denying the allegations as "completely unfounded."

The fake site in question calls itself "Reiwa Shinsengumi," and the three members each comment in support of Taro Yamamoto. Aoi Yu's comment read, "I sympathize with the policies of Reiwa Shinsengumi. Yamamoto-san is sincere in his dealings with young people. I myself would like to continue to work hard to contribute to society like Taro-san." It was as if the people themselves had made the comments.

This quickly spread on social media and aggregator sites, and Aoi's agency, Ito Company, released a statement on its official website denying the allegations, saying, "The comments made by Aoi Yu, a member of our company, about Reiwa Shinsengumi and Yamamoto Taro that were posted on a certain aggregator site today and spread on X are completely unfounded." Her husband, Yamazato Ryota of Nankai Candies, also denied the allegations.

Incent, the agency Yamamoto is affiliated with, also released a statement on its official website saying, "Yamamoto Mizuki has never made any comments supporting a specific political party."

Reiwa Shinsengumi also released a statement saying, "We have confirmed that there is a website that lists celebrities who support Reiwa Shinsengumi and their messages of support as if they were fact, under the title 'A summary of celebrities who support Reiwa Shinsengumi!' However, Reiwa Shinsengumi is not aware of the content posted on this website at all, and we believe that there is no truth to it," and said that they are considering legal action.

This time it looks like it will end up being just a small uproar, but needless to say, if it had happened during an election period, it would have had a huge impact on society.

Regarding the site that was created this time, the creator said, "It was created using AI," and continued, "I neglected to check that the comments made by the AI ​​were probably correct." It appears that he is claiming that he judged the comments made by the generating AI to be correct. It seems that the site creator himself had faith in the credibility of the AI ​​and would not lie. Of course, we need to be skeptical, but there may not have been any ill intentions.

Recently, a fake video of Prime Minister Kishida Fumio was made, causing a stir. With an office-style wall in the background, Kishida repeatedly made vile remarks, and the comments were displayed along with the logo and subtitles of Nippon Television's news program, making it seem like the real thing at first glance.

This video was spread on X (formerly Twitter) and other platforms, and even recorded over 2 million views, causing a huge reaction.

In an interview with a TV station, the creator said, "I know it sounds like I'm getting angry, but is it really that big a problem? They call it fake news, but they make it obvious that it's a joke, and if that's not allowed, then aren't all cartoons and other things not allowed?"

In addition, an advertisement video in which a Nippon Television announcer called for participation in high-yield investments on a news program was spread on social media, and Nippon Television warned that it was "altered." The fake advertisement, "100,000 yen in 3 days," was actually an altered news video from August this year about the "obligation to make an effort to wear bicycle helmets."

In the past, a video of President Zelensky calling for "surrender" was spread on Russian social media. Images of former US President Donald Trump desperately trying to escape from police officers surrounded by him were also featured in the media.

At the time, Trump was facing possible indictment over the allegations of paying "hush money" to porn actress Stormy Daniels. In fact, this was a fake image, which Elliot Higgins, founder of the British investigative journalism agency Bellingcat, confessed to creating using an AI tool called "Midjourney" and posting it on X.

Fake photos and videos created using AI are called "deepfakes." In the past, pornographic videos that synthesized female singers and actresses such as Taylor Swift and Scarlett Johansson have caused problems. In 2018, a fake video of former US President Paltrow Obama insulting Trump also went viral.

Social media companies such as Meta and X, which operate Facebook and Instagram, and TikTok, have said that they prohibit deepfakes on their platforms, but the technology to detect them has not yet been developed, and they are already spreading before they can be found.

Advances in generative AI have made it very easy to create deepfakes that are more realistic. Image creation and video editing programs incorporating generative AI have further evolved, and as long as you have the original image or video, you can easily create photos that show the person in them or videos that make it look like they are speaking.

We are now in an age where anyone can easily create and distribute large amounts of "fake news" with generative AI.

Last year, someone posted comments on Reddit, a major US bulletin board site, at a rate of one per minute for a week. The bot interacted with dozens of users before it was discovered to be GPT-3. Someone created a bot using GPT-3 and posted comments on Reddit. It is said that if the bot had used GPT-4, the bot's identity would not have been revealed.

Generative AI can easily create fake news and fake comments. The result is a phenomenon called an "echo chamber," in which certain opinions and beliefs are amplified in a closed space and other opinions are not accepted.

The US Capitol siege on January 6, 2021 was caused by conspiracy theories amplified by echo chambers. Supporters of Trump attacked the Capitol, claiming that there was fraud in the 2020 US presidential election. The attackers were followers of the QAnon conspiracy theory.

By misusing generative AI, such conspiracy theories can be easily created, and the images and videos that serve as the basis for them can also be easily created, so if they are spread on social media using AI bots, anyone can easily be brainwashed. Even if it does not go as far as brainwashing, it can be used to influence public opinion.

In addition to deep fakes, generative AI can also be misused for autonomous driving bombing terrorism, automated phishing scams, and AI hacking of corporate networks. In military use, autonomous drone bombing, misuse of facial recognition, and manipulation of the stock market and cryptocurrency market are also possible.

It would also be possible to use generative AI to leave evaluation comments on review sites, identify addresses based on photos, and easily counterfeit paintings, illustrations, music, anime videos, etc. For paintings, fake NFT works created by AI are already on the market. In addition, collections of beautiful girl images using AI are already being sold on Amazon and other sites.

In music, counterfeit songs that have been trained to learn Bach's music are being sold. In the future, there is a great possibility that counterfeit content of popular Japanese singers and idols will also be sold. In fact, a new song titled "Arakara" had been created by "AI Misora ​​Hibari," and although the rights issues had been cleared, there were also criticisms that "this is a desecration of the dead."

It is clear that in the future, there will be people who create content using AI, ignoring copyrights, and if the case goes to court, they may be found not guilty if they claim that "the AI ​​created it on its own."

In the fashion industry, AI is becoming more widespread in determining authenticity. US software manufacturer Alitheon has even developed technology that uses AI to determine whether something is real or fake using an iPhone app. However, according to the creators of such programs, there are also cases where AI intentionally determines that a fake is real.

Technology is also emerging to determine the authenticity of deepfake images and videos. The National Institute of Informatics (NII), part of the Inter-University Research Institute Corporation, Information and Systems Research Organization, has developed a program called "SYNTHETIQ VISION" that automatically determines the authenticity of fake face videos created by AI. It is currently being used by Cyber ​​Agency and is currently being used to detect deepfake videos of celebrities, etc. However, as AI itself continues to undergo deep learning and the technology becomes more advanced to create more sophisticated and realistic things, we will soon reach a point where human-made technology will no longer be able to keep up.

New laws and ethics regarding AI are required. Without these measures, we are now at a phase where what Elon Musk calls "dark AI" could dominate humanity.

Image: Created by Iolite, Shutterstock

Profile

◉ Akira Satomi
Former weekly magazine reporter. In charge of scandals and economic articles. Currently a freelance journalist, he writes for monthly and weekly magazines and online media. While he writes about crypto assets and other Web 3.0-related articles, he also writes about B-grade gourmet food such as ramen and izakayas.

Related articles

The cryptocurrency market is riding on misinformation: what will the media be like in the future and what will be its purpose?

"Grok" takes on OpenAI and Google in AI war: Elon Musk's two "weapons"