Princeton University computer science student Edward Tian created the GPT Zero program, which analyzes statistical criteria to assess whether a text was produced by a human or an AI.
Tian claims that the program evaluates the language for even, consistent complexity. This is a property of AI-generated literature, he noted, whereas human writers are more inclined to integrate variances into their works. The computer scientist provided a number of case studies in which his method was proven to be correct.
GPT Zero is now quite popular, perhaps because many people want to see if they have machine-generated texts or human-created works. GPT Zero is available via the developer’s website as well as Streamlit.
They attempted to test various texts generated by Chat GPT several times, but the results were mediocre at best. GPT Zero, for example, recognized lengthier messages written by Chat GPT for broad, lexical knowledge with some confidence. Short text recognition, on the other hand, failed. According to the developer, at least ten phrases must be examined in order for an assessment to be possible.
The International Conference on Machine Learning (ICML) recently stated that while writing scientific publications, writers should no longer utilize AI tools such as Chat GPT. The ICML, on the other hand, only outlaws writings “fully created” by artificial intelligence. The organizers clarified that they do not limit the use of tools like Chat GPT “for editing or enhancing content written by authors”. Because of worries that kids were utilizing the AI technology, the New York City Department of Education prohibited access to Chat GPT on its networks.