News

ChatGPT-generated content recognition systems are easily fooled

Delving deeper, the researchers found that some articles were likely classified as AI-generated content due to altered text. It was previously known that non-English speaking people use fewer words in their English texts because their vocabulary is considered much smaller compared to people who speak English, and artificial intelligence content detector programs consider such texts as artificial intelligence.

This means that if you use more literary text, these tools will not consider it as AI-generated content.

ChatGPT and literary language

Stanford researchers conducted another experiment and used artificial intelligence to determine whether the content type detection software correctly identified the intended texts as AI content or not.

A team of Stanford researchers used ChatGPT to generate responses to US college admissions essay prompts. Articles created with the OpenAI chatbot were reviewed by several content source detector tools. The research team receiving the software in question marks on average 70% of the mentioned articles as artificial intelligence. However, it eventually became clear that if you change the text of the article to more literary language, they will no longer be recognized as artificial intelligence.


Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button

Adblock Detected

Please consider supporting us by disabling your ad blocker