Ban ChatGPT? The educational and academic circle is “a little flustered”

  A well-structured academic paper may only take a few minutes to complete, do you believe it?
  Believe it or not, this is already a reality. ChatGPT, an AI tool released by the American artificial intelligence laboratory OpenAI in November last year, has rapidly become popular around the world at a jaw-dropping speed, because it can not only answer questions, but also let the human brain feel like creating poetry, writing code, and writing papers. Complicated things become easy.
  As its popularity soared, doubts and controversies followed, especially in education and academia, who were worried about the possible abuse of this technology, and even began to issue bans to build a “firewall”.
“Catching a knife” essay assignment raises concerns

  These days, ChatGPT is popular in overseas education circles, and many students quickly discovered the benefits brought by this artificial intelligence tool. Using it to complete homework and papers can reduce the death of brain cells and sleep more. The current situation can be described in one sentence: students enjoy using it, and teachers are “hard to guard against”.
  According to an anonymous survey conducted by the Stanford University campus media “Stanford Daily”, about 17% of the interviewed Stanford students (4497) said that they had used ChatGPT to help them complete their autumn assignments and exams.
  According to overseas media reports, assignments written by ChatGPT have become a phenomenon in American universities. According to a survey of 1,000 students over the age of 18 by online course provider, more than 9 out of every 10 students know ChatGPT, more than 89% of students use ChatGPT to complete homework, 48% of students Complete quizzes with ChatGPT, 53% of students use ChatGPT to write essays, and 22% of students use ChatGPT to generate essay outlines.
  Last December, Darren Hick, an assistant professor of philosophy at Furman University, gave students a 500-word post-class essay. It was obviously a non-difficult assignment, but he found that some students used ChatGPT to complete it. Darren Hick said that the free and immediacy of ChatGPT, as well as its super natural language processing ability, have caused many students to move their “brains”. They will hand over the assignments related to writing to ChatGPT, which is almost impossible for ordinary people. Distinguish the difference between articles written by real people and machines. He also said: “Maybe in the upgrade and optimization after ChatGPT, these traces will be more difficult to find.” The powerful AI generation makes Darren Hick feel extremely scared.
  As The New York Times reported in mid-January, Antony Aumann, a philosophy professor at Northern Michigan University, was grading a class on world religions he taught when he read a “best paper” free of typos and grammatical errors, with concise passages , appropriate examples and rigorous arguments surprised the professor. Unexpectedly, it turned out that this thesis was completed by the students using ChatGPT. In fact, some college students in the United States have used ChatGPT to write papers and won the first place.
  ChatGPT also showed amazing ability in some exams. A December 2022 paper published on medRxiv, a medical preprint database, showed that ChatGPT came close to passing the US Medical Licensing Examination (USMLE) in an experiment that the average American medical student needs four years of medical school and two years clinical rotations to pass this exam.
  Last month, Christian Terwiesch, a professor at the Wharton School of the University of Pennsylvania, published an article titled “Will ChatGPT Get a Wharton MBA?” ” article about testing ChatGPT on an exam in one of the core courses of its MBA program, and found that ChatGPT obtained a stable score between B and B-, and surpassed most students in the course.
  Students from many prestigious American schools use ChatGPT to “catch the knife” instead of writing papers or completing exams by themselves, and have performed at the top level. This means that ChatGPT can already engage in elementary and even higher-level academic research.
  A research report on ChatGPT was previously published in the American journal Financial Research Letters. The report shows that it can independently write a financial paper, and the content of the paper can be accepted by relevant academic journals.
  Sure enough, in addition to the teaching process, ChatGPT has also appeared in scientific literature in academia, and even appeared in academic papers as an author.
  For example, in an article published by Insilico Medicine, a Hong Kong artificial intelligence drug research and development company, in December 2022, ChatGPT was listed as the first author in an article using ChatGPT to speculate on the application of rapamycin, which caused controversy in the industry. ChatGPT is also listed as an author in the aforementioned paper published on the medical preprint database medRxiv that studies the performance of ChatGPT on the US Medical Licensing Examination.
  For another example, on December 27, 2022, Catherine Gao et al. from Northwestern University published a research paper on the preprint database bioRxiv. The research team used ChatGPT to write a convincing abstract of the research paper, with an original score of 100. %, and passed the plagiarism check, even human scientists have difficulty distinguishing the true from the false.
  According to the news department of the authoritative academic journal “Nature”, as of now, ChatGPT has appeared as a co-author in at least four published papers and preprints.
  Writing papers with AI is too easy. According to a survey conducted by “Nature” in December last year, 20% of the 293 interviewed professors and teachers have discovered or witnessed students using ChatGPT to complete assignments or papers. Regarding this phenomenon, Stephen Marche, a Canadian writer, appealed bitterly: University papers are dead!

  Students from many prestigious American schools use ChatGPT to “catch the knife” instead of writing papers or completing exams by themselves, and have performed at the top level. This means that ChatGPT can already engage in elementary and even higher-level academic research.

  At the same time, many scientists began to worry about the possible academic chaos caused by ChatGPT entering the academic circle. Although most of the papers that list ChatGPT as the author involve artificial intelligence, and even the content itself is related to ChatGPT, it still arouses concerns and controversies in the academic community.
The educational and academic circles resort to the big move of “blocking”

  In the face of ChatGPT’s “making waves” and the resulting concerns and controversies in the academic community. Top academic journals such as “Nature” and “Science” finally couldn’t sit still, and began to introduce restrictions on the use of ChatGPT.
  ”Science” recently published an article titled “ChatGPT is interesting, but it is not the author”, clearly stating that the use of text generated by ChatGPT in submitted papers is prohibited. Because “Science” requires that submitted papers must be original, it is not possible to plagiarize the text generated by ChatGPT, let alone list ChatGPT as the author of the paper.
  Holden Thorp, the editor-in-chief of the “Science” series of journals, publicly stated that the academic articles published in “Science” and its subsidiary journals are not allowed to list AI as the author of the paper. Using AI-generated text without proper citations could be seen as plagiarism, he said. At present, the “Science” series journals are updating the corresponding license and editorial policies to clearly state that text generated by ChatGPT (or any other artificial intelligence tool) cannot be used in the work.
  ”Nature” stated in the updated submission rules that the text generated by the large language model can be used in the paper, which also includes ChatGPT. But the large language model is just a tool, which should be properly introduced in the method section of the paper, and ChatGPT cannot be listed as the author.
  On January 3, the International Conference on Machine Learning (ICML) issued a call for papers for 2023. The conference requires all researchers, including reviewers, to prohibit the use of large language models (LLM) to write papers, such as ChatGPT.
  Journal editors, researchers and publishers around the world are currently debating the place of such AI tools in the published literature, and whether it is appropriate to list robots as authors of papers. In an interview with The Guardian, academic journal publisher Elsevier expressed similar views to Nature.
  According to the opinions of multiple publishers and preprint platforms cited by the News Department of Nature, because they cannot be responsible for the content and integrity of scientific papers, artificial intelligence such as ChatGPT does not meet the standards of paper authors. Some publishers, however, say AI contributions to other aspects of paper writing can be recognized beyond the question of authorship.
  ”In terms of informed consent for the submission of papers, responsibility for papers, and paper revisions, I think the current AI cannot meet the requirements.” Zhang Sen, who is studying for a doctorate at a public research university in Michigan, was interviewed by China Jiemian News Said, “This is a relatively fixed set of rules formed after countless games in the academic community for hundreds of years. Since AI is currently unable to participate in research work under this set of rules, I think it is more appropriate to list AI’s contribution. Include in the statement at the end of the paper.”
  According to the standards of the International Committee of Medical Journal Editors (ICMJE), four conditions must be met at the same time to be identified as the “author” of a paper, including substantive work on the conception or design of the paper; or interpretation of data; drafting or in-depth revision of the important intellectual content of the paper; confirmation of the final version and informed consent to the submission of the paper; and agreement to be responsible for the content of the work to ensure that responses to surveys and issues related to the accuracy and integrity of the research work are resolved.
  In addition to the restrictive rules put forward by the academic circles, the educational circles have also initiated “blocking” measures against ChatGPT.
  The Department of Education in New York, USA also discovered some harm that ChatGPT may bring, and officially promulgated the “ChatGPT Ban” on January 3. Neither teachers nor students can use ChatGPT on the network and equipment of public schools in New York City. .
  In response to the repeated behavior of students using ChatGPT to complete assignments and papers, Stanford University recently launched a “counterattack” against ChatGPT. Stanford researchers have launched DetectGPT, aiming to be one of the first tools to combat machine-generated text in higher education. DetectGPT is also an AI tool that can detect a specific tendency to generate text based on a large language model, so that it can determine whether the text is generated by AI.

  Zhao Xing believes that the opportunities brought by ChatGPT technology outweigh the challenges. For education and academia, instead of panic, it is better to think: Why are we defeated by machines again and again?

Is artificial intelligence against artificial intelligence a good way?

  Zhao Xing, chief expert of the Metaverse and Sapiens Laboratory of East China Normal University and director of the Metaverse and Virtual-Real Interaction Joint Research Institute, told Xinmin Weekly that there are currently some evaluation and inspection methods to detect papers that use ChatGPT. The idea is to use artificial Intelligence against artificial intelligence. At present, there are already applications on the market, and even OpenAI, the publishing company of ChatGPT, has begun to launch AI detection tools, such as OpenAl GPT-2 Detector, and others include Text Analysis Toolkit, etc. The content generated by Life Intelligence has traces to follow, and it is technically feasible to perform content detection.
  In addition, Universities UK said it was monitoring the issue closely but had not yet actively addressed it. TEQSA, Australia’s independent higher education regulator, said institutions needed to clearly define their own assessment rules and communicate the information to students.
  The reporter asked several colleges and universities in Shanghai, and they all said that they have not yet issued restrictions on ChatGPT, which may be related to the conditions of use of ChatGPT in China. However, Zhao Xing revealed that many domestic academic journals and scholars are beginning to discuss and study this topic.
Face up to the challenges and progress of technology

  A series of “blocking” has proved the strength of ChatGPT to a certain extent. However, whether the education and academic circles must regard ChatGPT as a “scourge” and believe that it has all kinds of harm but no benefit, many experts and scholars in the industry have also expressed different views.
  Some scholars said that in the higher education industry with a high knowledge threshold, it is also widely used in multiple scenarios. If ChatGPT is properly used in the correct scene, it may also improve the productivity of scientific researchers.
  For example, quickly retrieve relevant literature in the field: ChatGPT can quickly retrieve relevant literature in a specific field, or search for related research directions for a certain topic, which will be helpful in the literature review of all papers and the exploration of research directions for researchers. beneficial. In the process of reading the literature, you may encounter a long content. At this time, you can also use ChatGPT to summarize the main points. In addition, based on the language requirements for journal publication, ChatGPT can also guide writing, polish grammar, help authors reduce grammatical errors in paper writing, and optimize text expressions.
  Of course, although using ChatGPT can quickly generate answers, and even exceed expectations to respond to your needs, it cannot guarantee that the information is correct, and technical ethics issues are also open to discussion, and limitations still exist.
  Terwiesch, a professor at the Wharton School of the University of Pennsylvania, said he was in awe of ChatGPT’s user experience and the answers it gave. However, he also pointed out that ChatGPT made major mistakes in some fairly simple situations, showing a ten-fold deviation in the answer of a question, even below the academic ability of a middle school student.
  ”We have many reasons to believe that technology is getting better over time. But for many complex problems, we still need human participation.” Professor Terwiesch believes that as an educator, we should continue to teach students basic knowledge And skills, which are the prerequisites for understanding complex problems in the future.
  Professor Zhao Xing also pointed out that ChatGPT also has certain defects in solving problems, such as cutting-edge, innovative, logical, etc., and there are still significant deficiencies. The more core knowledge creation is, the more powerless AI is. The current popularity of ChatGPT has an element of freshness. Real academic research and innovation still require real down-to-earth scientific exploration.
  He does not think that tools such as ChatGPT will have a great impact on the education industry in the short term. “The process of education is not only the exchange of knowledge, but also the shaping of human nature in the communication with people. ChatGPT uses a large number of manual annotations, even It can be ‘pandering’ and ‘greasy’, but it can’t erase the fact that all interlocutors know it’s not human”. Zhao Xing said.
  In history, education and academic research are the two legs that support technological breakthroughs, but this time the industry achieved a breakthrough in ChatGPT technology, but hit the roots of education and academia, which is why the academic and higher education circles are highly vigilant . Pan Enrong, a professor of the School of Marxism at Zhejiang University, wrote in the “China Science Daily” that the importance of the relationship between “human and machine” has increased, and it will rise to an equally important position as the relationship between “human and human” in the future. It is precisely because of such changes in the underlying logic that the educational and academic circles panicked because of “misjudgment” about artificial intelligence and its social impact. He said that human and machine equal rights and co-evolution. We should recognize and face up to the capabilities of artificial intelligence.
  Zhao Xing believes that the opportunities brought by ChatGPT technology outweigh the challenges. For education and academia, instead of panic, it is better to think: Why are we defeated by machines again and again? He said: “Fighting against AI evolution is not as good as human beings evolving in a faster way. Education has not been able to make human beings evolve quickly. This is a problem we need to reflect on.”

error: Content is protected !!