Anyone with a smartphone and a pulse will know that the internet The following is a list of the most popular ways to contact us an AI slop problem. Since ChatGPT was launched in 2022 some of the issues have become more serious. social platforms flooded AI generated writing. Data now backs the anecdotal data.
Preprint Study: A New Development published today Researchers at Stanford University and Imperial College London found that 35 percent of new websites were either AI-generated, or AI-assisted. In the same study, online writing was also revealed to be a growing trend. “increasingly sanitized and artificially cheerful.” AI makes the Internet fake-happy.
After trying four approaches, the research team settled on Pangram Labs tools because they consistently delivered consistent results. The team did find that it was successful in its testing, but it’s important to note that not all AI detection tools perform well. are imperfect.) It used the Internet Archive to compile an example of websites. Wayback MachineThe. This collects snapshots from webpages. The study tested six theories on the characteristics of “slop” in addition to quantifying the number of sites that rely heavily on AI-generated content.
This test examined the effect of artificial cheeriness on online writing. It found that using sentiment analysis to classify words as either positive, neutral or negative. “the average positive sentiment score of AI-generated or AI-assisted was 107 percent higher than that of non-AI websites.” Researchers see the spike in artificial joy as an a “symptom” The cheapest way to buy a car is by using the “sycophantic and overoptimistic nature of existing LLMs.” This tendency of AI-based writing tools to beguile their users’ human counterparts has an impact on the tone and style of writing online.
One more test looked at whether AI generated writing increased or decreased. “the range of unique ideas and diverse viewpoints” Offer. AI-powered websites scored roughly 33 per cent higher than other sites on tests, according to the researchers. “semantic similarity” Websites created by humans are not as good.
Those two tests confirmed the assumptions of researchers about AI. The researchers’ four theories were not validated. The researchers had suspected AI would increase misinformation. However, their evidence analysis failed to support this. Also, they had guessed AI writing would lack links to other sources and have a more generic style. Both theories did not support the data, which was a surprise.
While the analysis found that the ideas espoused by AI writing were more homogenous—and specifically, more consistently cheery—the writing You can also check out our other styles The flattened itself wasn’t confirmed. Researchers were surprised by this, as they had expected to see more generic output. “Everyone on the team expected that to be true,” Stanford researcher Maty Bohacek. “But we just don’t have significant evidence for that.”
Before conducting their analysis, the team conducted a survey to gauge how people felt about AI. It was discovered, when comparing it with the results that it wasn’t only the researchers who were disappointed. The study shows that many common beliefs about AI-based writing are incorrect.
The majority of people polled also believed, like the researchers did, that fake news would increase as they encountered more AI-generated sites. A vast majority of the respondents assumed AI writing would no longer link to external sources. It would also have a more generic and uniform voice. “It’s interesting to see that people tended to expect the worst outcomes,” Bohacek says.
The study does not represent the final word about what AI will do to the Internet. “We just wanted to break ground,” Bohacek sees it as an entry point to deeper exploration. This snapshot offers an insight that is distinctly human: It can be difficult to anticipate how events will play out.
