According to a recent survey by the "news credibility rating" website NewsGuard, a number of AI chatbots will quote the "Pink slime" website multiple times.
"Pink meat dregs" is actually a product of the old times. This means that at first glance, these websites may look like ordinary local news media, but in fact they may be supported by groups with political or ideological purposes, without clearly revealing the background, and are as pulpy and mixed as processed meat.
The problem behind this is: when users use AI tools to ask questions about candidates or political issues, these already existing sources of information begin to enter the AI’s search, citation, and summary process.
This brings to mind another problem that has existed in Taiwan for many years: content farms. Websites such as “content farms” have several common characteristics:
1. Extensively rewrite and collage existing content from other media or the Internet
2. Attract clicks with exaggerated titles
3. The author or editor, or the source of the interview process is not clear
4. Produce a large number of articles in multiple fields in a short period of time
5. Mass distribution through Facebook, LINE, search engines and other channels
6. Like news websites but lack the correction or censorship mechanisms of news organizations
Although "pink meat scum" and "content farms" as they are often called in Taiwan are not synonymous, they overlap: both may produce and rewrite content in large quantities, and the background sources are opaque. What needs to be clarified is: content farms are not completely equal to fake news, and articles may be mixed with true and false. More often than not, existing information is re-collaged and rewritten, and then added with exaggerated titles.
Therefore, the question not only includes whether the content is true or false, but also "who produces this information? Why produce it? Where does the original information come from? If the content is wrong, who is responsible?"
The operation model investigated by "Reporter" in 2019 for "Secret News" is one of the cases. It copies and rewrites existing news in large quantities, and has a considerable amount of dissemination on Facebook. The reporter mentioned that according to information from the "FB Page Dashboard", "Secret News" became the most shared domain by Facebook users in a single week in April of that year, with the number of shares being nearly five times that of "Liberty Times".
In the past, the entrance to content farms was Google or Facebook; now it may be AI. In the past, you might directly read about content farms, and after browsing the website, you knew it was a content farm; now you might read about content farms, but you don’t know it comes from a content farm.
"Pink scraps" and content farms are actually not new products, but new questions have indeed arisen in the AI era: When we hand over part of the work of "searching for information and organizing information" to AI, will these sources of information that originally have problems also be sorted into answers?
So there are actually two things to do. First, advance warning. When erroneous information enters the subconscious mind, the correction effect of refuting the rumor afterwards may be very limited. If either the national security unit or the civil defense organization detects suspected coordinated behavior by a large number of cyber forces, the best way is to provide early warning before rumors take shape. If we take the recent Trump-Xi meeting as an example, the White House immediately released a "fact list" at the White House on the day the Trump-Xi meeting ended, clearly listing what was discussed.
Second, list your sources. This is probably the most important step when we use AI tools. When we use ChatGPT and other AI to check news, politics and public issues, in addition to asking "What is the answer?" we also need to ask: "Where is your source of information?" And don't stop at the source name listed by the AI, but really click on it to see: "What kind of website does it cite?"
