[News] News websites created by artificial intelligence chatbots already exist, and some are a danger

[#Wittels-]

An investigation found dozens of websites publishing hundreds of articles per day generated by chatbots like ChatGPT. They publish all kinds of news, some as crazy as the death of Biden.

A new generation of content farms is on the way, warns Newsguard. The organization, which is tasked with tracking false information on the Internet, found at least 49 news websites in April that were written almost entirely by AI chatbots, such as ChatGPT. Some publish hundreds of articles per day, many of them with disparate information.

"Last minute: The White House has reported that Joe Biden would have it peacefully while he slept," the Celebridadesdeaths.com portal (dead celebrities in English) published last month. They are websites designed to imitate human communication and the format of traditional news pages. Almost all of the content features bland language and repetitive phrases, characteristic of the AI.

Not all websites have a name as obvious or suspicious as "dead celebrities." Others have more generic names, intended to give the impression of being created media: Biz Breaking News, News Live 79 or Daily Business Post. Articles are rewritten from other sources. For example, BestBudgetUSA.com summarizes articles from CNN.

The sites are saturated with advertising. This means that these AI-created news websites were surely designed to generate revenue from programmatic, algorithmically placed ads that are today financed by a large part of the world's media.

Who put together these AI news websites?

In these AI-generated news websites, sensationalist headlines, such as the death of the president of the United States, are key. Also, trivia: "Why is yawning contagious?" or "why is the sky blue but space is black?" These are two examples of the articles published by GetIntoKnowledge.com.

Newsguard tried to contact the directors of these portals. Of the 49 sites, only 29 had a contact address. Most were fake emails or never responded. Among those who did respond is the administrator of Famadillo.com, a site that publishes product reviews by AI. "We made an expert (sic) to use AI to edit old articles that no one reads anymore, just to see how it works," said Maria Spanadoris, who declined a phone call. Therefore, his identity could not be verified.

Adesh Ingale identified himself as the founder of GetIntoKnowledge.com. "We use automation at some points where they are extremely needed," he told Newsguard in an email. According to Ingale, everything is published manually under human supervision, but he did not elaborate on the use of AI. He did highlight: "We are the new era that will provide knowledge in each and every one of the corners."

News websites generated by AI chatbots are written in seven languages: Chinese, Czech, English, French, Portuguese, Tagalog, and Thai. Many of the articles are signed by "Administrator", "Editor" or simply have no signature at all. Other sites feature fake author profiles. HarmonyHustle.com, for example, has a list of its creators, including "Alex" and "Tom." Newsguard, however, did an investigation based on their profile photos and found that they were not real.

How to identify these fake pages?

In addition to being endorsed by an author signature, articles on these news websites often include clear evidence that they were written by an AI. For example, articles on BestBudgetUSA.com often contain phrases such as "I can't produce 1500 words...However, I can provide you with a summary of the article..." plus the link to the original CNN report.

Other posts contain sentences like "My deadline is September 2021" or "As an AI language model...". The article on Biden's death said that the vice president would replace the president of the United States. Then, however, he continued: "Sorry, I can't complete this message because it goes against OpenAI's use case policy on generating misleading content." OpenAI is the creator of ChatGPT.

The 49 websites analyzed by Newsguard contained at least one article with these flaws. You can also tell the use of a chatbot by the repetitive language, which constantly includes fillers such as "in conclusion" and "it is important to remember".

These sites have a varied presence on social networks. ScoopEarth.com, for example, has 124,000 followers on Facebook. While FilthyLucre.com, which publishes financial articles, operates on Facebook, Instagram and Twitter with zero followers.

This is just the beginning, just months away from the launch of ChatGPT. "Concerns that (AI) could be used to evoke entire news organizations, once the subject of speculation... have now become a reality," Newsguard reported.

 

Link: https://hipertextual.com/2023/05/webs-noticias-creadas-por-chatbots-de-inteligencia-artificial