Concerns Over Copying Assignments with ChatGPT Spur Busy Development of AI Detection Technology

Smart ChatGPT... Concerns Over Misuse for Cheating
GPTZero and DetectGPT AI Detection Technologies Emerging Rapidly

[Asia Economy Reporter Yoon Seul-gi] As OpenAI's chatbot ChatGPT gains worldwide popularity, concerns are growing in the education sector that it could be used to write assignments on behalf of students. However, AI detection services such as GPTZero and DetectGPT, which can identify whether AI has been used, are also evolving.

ChatGPT is a large-scale language model (LLM) developed based on GPT-3.5, an improved version of the super-large AI (artificial intelligence) GPT-3. Although the number of parameters of GPT-3.5 has not been disclosed, GPT-3 had 175 billion parameters. Parameters act like neurons in the human brain, and the more there are, the better the performance.

ChatGPT has learned from a vast amount of text to generate consistent and diverse text. Recently, it made headlines by passing the MBA graduation exam at the prestigious Wharton School of the University of Pennsylvania, achieving passing scores on the medical licensing exam, and receiving an average grade of C+ on the law school graduation exam at the University of Minnesota.

The photo is not related to the specific content of the article. Photo by Pixabay [Image source=Pixabay]

Its ability to produce results that are difficult to distinguish from those of humans is a major advantage, but there are concerns about potential misuse. There is a possibility of cheating using ChatGPT for assignments or exams. In response, cities like Los Angeles and New York in the U.S., where the ChatGPT craze first began, have taken preemptive measures by banning access to ChatGPT on all networks and devices within school systems.

In South Korea, as ChatGPT gained popularity, alarm bells rang on university campuses ahead of the March semester. This is because students can use ChatGPT to write assignments or reports. There have already been cases of misuse; recently, students at an international school in the Seoul metropolitan area used ChatGPT to write English essays.

However, these students were all given zero points after their use of ChatGPT was detected. This was thanks to the application GPTZero, which can detect whether ChatGPT was used. GPTZero is a ChatGPT detection program developed by Edward Tian, a student at Princeton University, that can identify if an essay was written using ChatGPT. By inputting a file or text, it compares it with ChatGPT language data and calculates the probability of plagiarism. Tian also released an upgraded version of GPTZero on the 30th of last month.

A program called GPT Zero that can detect the use of ChatGPT. Photo by GPT Zero.

As text-based AI advances, AI detection services are also developing. Stanford University in the U.S. has launched 'DetectGPT.' DetectGPT is a model that detects text generated by GPT-3. It works by inputting the sentence to be analyzed, transforming words within the sentence into multiple variations, and then comparing curves to determine AI usage. After analyzing the curves of the original and transformed sentences, if the curve is irregular, it is judged to be human-written; if consistent, AI-written.

OpenAI also offers a free service called 'Classifier' that can determine whether text was created with the participation of AI such as ChatGPT. By copying and pasting the text to be verified into this app, it diagnoses whether AI wrote the text.

Meanwhile, as AI has advanced to a level that causes fear and the potential for misuse has increased, OpenAI has mentioned the need for regulation. Mira Murati, OpenAI's Chief Technology Officer (CTO), said in an interview with the U.S. news magazine Time on the 5th (local time), "The high popularity has caused some ethical issues," adding, "AI tools can be misused or used by bad actors."

She continued, "This has raised the issue worldwide of how to control AI like ChatGPT," and said, "It is important for companies like OpenAI to be controlled and to responsibly instill (regulation) into public awareness." She emphasized, "However, our staff is small, and much more information beyond technology is needed for this," adding, "We need help from regulators, governments, and everyone else."

This content was produced with the assistance of AI translation services.