Kansas, 15 June, 2023, (GNP): As students and professionals from various fields seek assistance from the well-known OpenAI ChatGPT chatbot, researchers from the University of Kansas have developed a method to detect artificial intelligence (AI)-generated text and written drafts.
The ability of ChatGPT to produce text and dialogues that resemble human speech recently caused quite a stir when it became known to the people.
However, a study published on Wednesday in the journal Cell Reports Physical Science found that there are only a few clear signs that can help humans distinguish AI chatbots from real people.
The team has used these factors to develop a system that, over 99% of the time, can identify academic research papers written by AI.
Also Read: AI worries Hollywood actors
The research placed a lot of focus on perspective articles, which provide a thorough summary of a particular study topic typically written by scientists.
To train the model, the team chose 64 points of view and produced 128 articles using ChatGPT that addressed the same study themes.
“To be exact, we built our strategy using just 64 human-written documents and 128 AI documents as our training data,” Heather Desaire, a chemist at the University of Kansas, said.
After comparing the writing, they discovered predictability, which is a distinguishing feature of AI writing.
Also Read: AI can decode human thoughts: Scientists
Despite ChatGPT’s similarity to human writing, humans are still different from AI because they have more complex paragraph structures and a range of sentence lengths, phrases per sentence, and phrase counts.
As a result, the analysis team created a synthetic intelligence text detector that was superior to a business one.
Moreover, they also intend to assess its application to larger datasets and different forms of scientific writing, focusing on resilience as AI chatbots get more sophisticated.
While the model differentiates between AI and people, the analytical team found that their model could be easily replicated and made by others.
