The new tool, named “AI Text Classifier” by OpenAI, was released on January 31 (US time). The San Francisco-based company describes it as “a fine-tuned GPT model to determine how likely a piece of text is to be generated by AI or by a human”.
After reviewing the text, OpenAI’s new tool will evaluate according to different levels to conclude whether the text is AI-generated or not.
“Our use when creating the ‘AI Text Classifier’ was to distinguish the difference between human-written content and AI-generated content,” said OpenAI Company – ” The results can be useful but are not the only evidence when deciding whether a text has to be AI-generated.”
ChatGPT launched on November 30, 2022, is a program combined with artificial intelligence to interact with people. The program uses a word-choice algorithm based on lessons learned from scanning billions of texts on the internet. From there, allowing the computer to chat and respond to questions from the user in real time.
Thus, people can ask ChatGPT to perform a series of tasks, such as gathering information, fixing broken computer code, creating Snoop Dogg-style rap lyrics, writing scripts for hypothetical scenes, writing scripts. nice essay asking “quote me a verse from Tale of Kieu” by great poet Nguyen Du…
The capabilities of ChatGPT have raised concerns among many educators. Even the New York City Department of Education – USA banned ChatGPT from school devices and networks in early January, citing concerns about “negative impact on student learning”.
That has also fueled OpenAI’s efforts to create AI-generated text detection tools.
However, OpenAI’s “AI Text Classifier” still has some limitations. Sample writing should be at least 1,000 characters, or about 150-250 words. OpenAI also acknowledges that the tool is not always accurate — AI-generated text can be edited to prevent detection tools and text classifiers from misidentifying even AI-generated patterns. and models written by humans.