The company responsible for developing Chat GPT, has recently launched a new AI text classifier called Zero GPT. This tool is designed to identify the origins of written content.
Chat GPT has gained significant popularity, particularly among students who have used the AI chatbot to produce convincing articles. However, the increasing use of Chat GPT has raised concerns about plagiarism. In response, OpenAI has developed Zero GPT, an AI text classifier that could potentially assist in identifying instances of plagiarism. Zero GPT has the ability to trace the creation of written content.
Chat GPT defines an AI text classifier as a machine learning algorithm that assigns text data into pre-established categories or labels. However, ChatGPT acknowledges that it did not have prior knowledge of the Zero GPT detector tool.
What is Zero GPT?
Zero GPT, described as an “advanced and reliable Chat GPT detector tool” on its website, operates similarly to online plagiarism checkers. To utilize the tool, users must upload text to the website, and ZeroGPT will then identify the segments that were “AI or GPT generated.”
Zero GPT exclusively analyzes AI-generated English text, and it provides a rating system that ranges from “most likely generated by AI” to “very unlikely.”
Chat GPT states that AI-generated text usually has distinctive characteristics that set it apart from human-generated text, such as repetition, incoherence, and the use of unnatural or ungrammatical language. A text classifier like Zero GPT can be trained on a large dataset of both AI-generated and human-generated text to recognize the patterns and features that differentiate the two.
Where Zero GPT Fails?
Despite its launch, testing of Zero GPT indicated that it was not entirely infallible. During a trial run, attempts to “train” Chat GPT using human-written articles as a reference resulted in inconsistent results from Zero GPT.
OpenAI acknowledges that the tool is currently imperfect. Therefore, the company recommends that caution be exercised when utilizing Zero GPT in academic dishonesty determinations.
According to Lama Ahmad, the policy research director at OpenAI, the tool should not be relied upon exclusively. Ahmad emphasizes that Zero GPT, like any other AI tool utilized for assessment purposes, is susceptible to error.
Recommended reading: Bard AI: A Deep Dive into Chat GPT Competitor