The Impact of AI-Generated Text on the Internet
The article discusses the challenges of determining the amount of AI-generated text on the internet. It highlights the difficulties in creating a representative sample of web content and the complexities of distinguishing between AI-generated and human-written text. The authors utilize various detection methods and conclude that Pangram v3 is the most effective in their tests.
- ▪Constructing a statistically representative sample of the internet is challenging due to the lack of a central index.
- ▪The authors use the Internet Archive's Wayback Machine and a multi-dimensional stratified sampling approach to analyze web pages.
- ▪They experiment with four AI text detection methods and find that Pangram v3 performs the best overall.
Opening excerpt (first ~120 words) tap to expand
How much new text on the internet is AI-generated? Answering this question is harder than it might seem. Constructing a statistically representative sample of the internet is difficult, as there is no central index, popular domains are vastly over-represented in most crawls, and archival coverage has shifted considerably over time. To work around this, we draw on the Internet Archive's Wayback Machine and apply a multi-dimensional stratified sampling approach, approximating a uniform random draw from publicly accessible web pages published between 2022 and 2025 (see Section 3.1 in our paper). On top of this sample, we need a reliable way to tell AI-generated and AI-assisted text apart from human-written text.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Github.
Discussion
0 commentsMore from Github
