No point in arguing about it – AI generated news is already here
NewsGuard published a report on 49 AI-generated news blogs, and TheMayor.EU sought to verify these claims.
Since the public release of Chat GPTfact-checkers and journalists have feared the day when content farms and troll farms start getting wiser Mass Adoption of Generative AI Language ModelsWell, it seems that day has come.
On May 1, NewsGuard, a company that provides accuracy ratings for news sites, 49 websites that only use AI, with little to no human editing. These are mostly low-quality clickbait articles on various topics such as politics, health, entertainment, lifestyle, finance, religion, technology, and more.
NewsGuard found the AI website works in seven languages: Chinese, Czech, English, French, Portuguese, Tagalog, and Thai. Some publish hundreds of articles a day.and some progress false story.
question
According to the report, all 49 websites have characteristics. programmatic advertising, Algorithmically placed. In this case, the content acts as a search engine optimization,You can get a lot of ad revenue with as little effort as possible.
our team The Mayor.EU I visited some of the listed websites and came to a surprising conclusion. Almost indistinguishable from regular content farms – In other words, nothing new on the Internet.
The websites listed contained all the hallmarks of a low-cost content farm you would expect. Article written by admin (rather than real personal names), no contact information or company details, bland language and dangerous English.
However, these factors are not necessarily indicative of AI-generated content. Websites like this have been a staple of the Internet for quite some time.And although their content always certainly has Felt very “robotic”, they were written by real people. This made me question NewsGuard’s claims that the content was actually generated via his AI language model. So I tried validation.
The hole in the AI news site is only inches deep
One of the big claims NewsGuard is making is that all 49 websites have At least one article covering the following AI error message: “Deadline is September 2021”, “As an AI language model”, “Can’t complete this prompt”, etc.
In fact, during our research, we found that some of the listed websites were clearly created by AI. Efforts to cover up that fact were minimal at best, and the supposed “deception” was only inches deep. It became like Manually remove error messages – This is still a very low bar to pass.
All of the websites we investigated had the characteristic look of a content farm, but the content was undoubtedly 100% AI and extremely difficult to identify and assert. I also found some. No human is required to apply. And maybe that’s the scary part. All of the listed websites were able to produce a huge amount of content.
Journey of Checkers
A website we visited GetIntoKnowledge.comThe website appears to have been active since December 18th, 2022 and stopped publishing new articles on April 23rd of this year. All are credited as authors to GET INTO KNOWLEDGE, covering topics ranging from engineering to biography. The site also has a section called “Podcasts,” which read articles highlighted by an artificially generated voice.
Also, the article What is a Leap Year? Why Do Leap Years Every Fourth? – Amazing Science Explained contains a famous phrase.As an AI language model, I’m excited to delve into the reasons behind leap years and try to make it understandable to everyone.“
Screenshot from Get Into Knowledge, Source: TheMayor.EU
can’t skip mention CountyLocalNews.comthere seems to be only one author – one John Ekpa. Since the website was first launched on February 7, 2023, Ekpa has made a remarkable With 8,390 articles, he is probably the world’s most prolific author.
I started writing slowly at first, writing only a few days, and by March 8th, I had written 14 articles. Just today, May 4th, John Ekpa has already managed to write as he of EET is ready to publish this article at 12:15. 36 articles!
The title of the article he published on April 20th: “Living in a muddy trench surrounded by fire in Russia Note: As an AI language model, I am not programmed to promote or endorse violence, hate speech, or discrimination.
Screenshot from CountyLocalNews, Source: TheMayor.EU
Another website that NewsGuard points to is Harmony Hustle.comCreated in April 2023, it covers topics from lifestyle to cryptocurrencies. Also, it does not have an About page and operates anonymously.Most of its authors have common names such as Tom, Leo, Nick, Sophia, However, they have profile pictures and we decided to track them down. InVid- An attempt to prove if these are in fact real people.
“Leo” grainy profile picture, Source: HarmonyHustle
His grainy portrait shots made it look like something was going on with his fingers, so I decided to start with Leo.one The main tell using AI-generated images is the finger. Usually they are broken or irregular numbers. However, a reverse image search revealed that he was just a stock photo person. shutter stock and ‘BigStockPhoto.com’ This photo is titled “Young man eating Asian sushi with chopsticks”.
“Leo” definitely looks like a real person, even though his hands are photographed in weird shapes. Also, just because this person is a stock photo model doesn’t mean he doesn’t write for Harmony Hustle.So we looked at sophia We found her on Unsplash, a stock photo model for creator Jonathan Borba.
Sophia’s grainy profile picture, Source: HarmonyHustle
Source photo from Unsplash, Source: Jonathan Borba
However, despite our best efforts, No articles found with AI error messages upon Harmony Hustle. As far as we are concerned, the entire website could have been written entirely by humans. It could also have been written by someone who deleted it.
Although the article ‘How artificial intelligence is revolutionizing the healthcare industry‘ is something Sophia supposedly wrote at least five times. The text contains very common and repetitive lines, but I haven’t been able to find 100% clear indications. In theory, it could be written or at least edited by a real person.
Additionally, the site owner could have cleaned up the content since the NewsGuard article was published. Low standards, but enough to cast doubt on the conclusions.
Still, all of the above It also highlights the need to find a way to unambiguously distinguish whether a piece of content is AI-generated or genuine. Ironically, it could be another job for AI.
