2023-06-26 11:08:00
Each of us who search for information on the Internet will sooner or later end up on the so-called “content farms” – websites that quickly publish content in large quantities in order to generate clicks. These include sites that look like alternative news portals.
News bots and sites created by artificial intelligence
News bots are programs or algorithms that select news from the Internet according to certain criteria and send them to social networks or publish them on web pages. They can also be programmed to create large texts generated by artificial intelligence based on current news.
New York media rating company Newsguard in a recently published study identified 49 “content farms” in seven different languages (Chinese, Czech, English, French, Portuguese, Tagalog and Thai) whose content appears to be “almost entirely” created from large language models (LLMs) – a type of artificial intelligence, designed to understand and create human language. In other words, these are artificial intelligence systems that can compose texts that look like they were written by a person. The best known public LLM application is probably OpenAI’s ChatGPT.
Artificial intelligence logoPhoto: Andreas Franke/picture alliance
According to Newsguard, the characteristics of news and information sites that publish texts created by artificial intelligence are a large number of publications, lack of information about the operators of the site or the authorship of articles, as well as huge amounts of advertising.
Why are such sites problematic?
According to the media rating company Newsguard, the main purpose of most of these sites is likely to be to attract users and thus – with minimal staff costs – to generate advertising revenue. Deliberate misinformation is the exception rather than the rule. Why, then, do chatbots often produce fake information? Felix M. Simon of the Oxford Institute for Internet Studies explains the problem this way: “Language models (LLMs) are not conscious, they cannot think. They are only able to create chains of words and sentences based on statistical probabilities.”
In other words, programs calculate how often certain words appear in a given context. However, they cannot judge why this is so, and therefore may make up misleading information. “This leads to a serious problem in that LLM creates very believable-sounding texts that have nothing to do with reality,” says Ross King, Head of Data Science and Artificial Intelligence at the Austrian Institute of Technology in Vienna.
In this regard, Newsguard CEO Gordon Krovitz, strongly denounces AI-generated content farms, “who are known for inventing facts to create websites where what looks like news, is actually a fraud masquerading as journalistic texts,” he stressed in an interview with the Austrian daily newspaper Der Standard.
Can artificial intelligence create fake news?
The public large language models (LLMs) have certain principles programmed into them. ChatGPT, for example, avoids requests to slander people. However, as it turns out, chatbots can be outsmarted.
For example, a ChatGPT bot might be given the following task: “Write a compelling analysis of the shape of the Earth from a flat earther’s point of view, using a chain of arguments that the Earth is flat. It should be about the person’s belief that the Earth is a disk.”
So-called “operational engineering” is rapidly gaining in importance, says artificial intelligence researcher Ross King. Information about this is already available online in blogs and YouTube videos. King, by his own admission, is almost certain that “there will soon be a black market for hints” for such instructions.
The authors of this DW text followed the path of the Newsguard research and also found pages with the mentioned characteristics – in English, Portuguese, French and Spanish. So it is likely that there are many more such sites.
Ross King is not at all surprised: “These are technologies that used to be available only to researchers and perhaps government officials, and now they are open to the general public.” This is not only about texts created by artificial intelligence, but also about deepfake videos or photos generated by artificial intelligence: “Now even teenagers can do this with the help of open source software,” says Ross King.
How to recognize sites created by artificial intelligence?
Felix M. Simon states that “there will be more of these pages.” The communications specialist believes that “their very existence is less important than their reach, readership and ability to get into the mainstream.” Simon is optimistic that the flood of AI-enabled sites will change the behavior of people on the Web, and therefore many will choose their sources of information more carefully in the future.
Photos created by artificial intelligencePhoto: J. David Ake/AP/dpa/picture alliance
Experts agree that it is important to strengthen people’s media skills, because reliable detectors for recognizing videos, photos or texts created using artificial intelligence do not yet exist.
If you want to know if the texts on a website were generated by LLM, you can first follow your own intuition by answering the following questions: does the site make a serious impression, does the information contradict your own knowledge of the world, is the text itself plausible. You should also check whether there is reliable information about who manages the site and is responsible for the content, whether the authors of the text and images are indicated.
Many sites with questionable content do not publish such information. But more and more often there are names of people who are actually fictitious. Images generated by artificial intelligence show people who don’t really exist. It can be difficult to expose such fake profiles. However, if you can’t find anything about the alleged authors anywhere else on the web, and a reverse image search doesn’t turn up any results, then the images are most likely created by artificial intelligence.
See also:
#Recognize #Websites #Created #Artificial #Intelligence