Recently, the code interpreter version of ChatGPT was officially released to Plus users, which means that no professional programming technology is required, only natural language instructions can be used to operate ChatGPT to achieve complex tasks.
However, ChatGPT has been controversial since its inception, especially the academic ethics issues it raises, making its application in academia not easy. How to make increasingly mature artificial intelligence better serve academia, we still have a long way to go.
AI-generated content raises a host of academic ethical questions, as well as AI detection software.
Due to the powerful computing power of ChatGPT, it has gradually become a powerful tool for students to write papers, which has also attracted the attention of the academic community. As a result, some public schools in New York, Seattle, Los Angeles and other places have banned students from using ChatGPT.
Many scientists disagree with AI as an author's name in the article, and the editor-in-chief of the journal Science has made it clear that "we do not accept AI as the author of our published paper, and text generated using AI may be considered plagiarism without proper citation."
The editor-in-chief of the journal Nature also emphasized: "No large language model (LLM) can be accepted as a byline author of a paper. This is because the author's name represents a responsibility for the research work, which AI tools cannot assume."
As a result, a series of AI detection software, including Turnitin, GPT Zero and Compilatio, have emerged and many university teachers have applied them to their daily teaching, with GPT Zero being the most famous.
GPT Zero, hailed as the most powerful AI counter-tool, was released in March by Edward Tian, a fourth-year student at Princeton University. This tool is designed to assist educators in detecting works generated by ChatGPT and has successfully secured investment. As a result, Edward Tian has also been hailed as the "nation's student enemy."
However, with only a few tweaks, AI-generated articles can pass the test, which calls into question the true validity of AI detection tools.
According to GPT Zero, its AI text detection tool uses deep analysis technology to determine the origin of text, with a detection accuracy rate of up to 98%. They also promised continuous improvement to reduce the error rate to less than 2%. But is this accuracy really credible?
Deborah Weber Woolf, professor of media and computing at the University of Applied Sciences for Technology and Economics in Berlin, Germany, and research teams from several universities evaluated 14 detection tools, including GPT Zero.
The research team used ChatGPT to generate two texts, each slightly modified to mask the fact that it was generated by AI. One was manually edited by researchers who rearranged sentences and replaced words, and the other was rewritten using an AI paraphrasing tool called Quillbot.
They found that while the tools were 74 percent accurate at identifying ChatGPT text, the accuracy dropped to 42 percent after the text was slightly adjusted. This means that students only need to modify AI-generated articles slightly to fool the detection tool.
Daphne Hippolyto, senior research scientist at Google's natural language generation, raised another concern. "If automated detection systems are used in educational settings, it is critical to understand their false positive rates, as falsely accusing students of cheating can have serious implications for their academic careers," she said. ”
As for ChatGPT, if it can cultivate advanced AI users, perhaps it can achieve harmonious coexistence with the education industry.
ChatGPT, an artificial intelligence technology, is rapidly developing and opening up to all users, especially among the student body. However, on the one hand, the controversy of academic fraud continues, and on the other hand, the perfection of AI detection tools still needs to be improved. In this context, how will ChatGPT affect human education? Should we consider it a friend and welcome its presence, or should we see it as an enemy and try to exclude it with detection tools?
Feng Jianjun, a professor at the Institute of Moral Education at Nanjing Normal University, believes that in the past hundred years of educational development, whenever a new technology appears, it will trigger social concerns and academic debates. Because of the intelligence of ChatGPT, there are even more concerns about whether it may replace teachers, or even education.
He thinks this worry is like asking: When we have a car and learn to drive, do we still need to learn to walk? When we have computers, do we still need to learn mathematics? However, these are only tools used by humans, and they do not change humanity itself.
In the face of the emergence of ChatGPT, we cannot adopt an attitude of blind exclusion. The progress of technology is an inevitable trend of social development and cannot be stopped. The future of education is neither entirely human-led nor entirely dominated by ChatGPT, but rather by human participation with ChatGPT. Therefore, we must seek the symbiotic coexistence of humans and ChatGPT.
Kristen DiCerbo, chief learning officer at Khan Academy, once said, "GPT-4 is breaking new ground in education. Many people have been looking forward to this technology for a long time. This is a transformative technology that we plan to test responsibly to see if it can be used effectively for learning and teaching. ”
While actively exploring the potential of AI applications in education, we must also be cautious about the ethical issues that may be touched upon. To this end, we need to introduce regulatory technology and establish and improve the ethics, risk assessment and governance system for ChatGPT applications. Along the way, we need to reach consensus on certain key areas, including whether ChatGPT should be given authorship and how to assess its workload.
Rephrased into original copywriting:
While actively exploring the potential of AI in the field of education, we must also pay close attention to the ethical issues that may be involved. Therefore, we need to implement regulatory strategies to build and optimize the ethical standards, risk assessment and management systems of ChatGPT in practical applications. In doing so, we need to develop consistent rules in specific areas, including whether ChatGPT should be given authorship and how to evaluate the amount of work it produces.