Shulou(Shulou.com)11/24 Report--

February 22 news, in station B, there is UP master shared his own ChatGPT took only half an hour to write undergraduate thesis, successfully passed a domestic duplication website inspection, duplication rate is less than 9%. However, it was also found that most of the references listed in ChatGPT's papers were "fabricated" and did not exist at all.

According to a survey conducted by Study.com in January of this year among 1000 students over the age of 18, 48% of students completed quizzes with ChatGPT and 53% wrote essays with ChatGPT.

ChatGPT's amazing performance in writing format papers, passing professional examinations and other fields makes students cheer, but also makes universities and research institutions take the lead in worrying and vigilance."Anti-ChatGPT" takes the lead in becoming a new trend in academic circles.

Recently, the University of Hong Kong issued an internal email to faculty and students stating that ChatGPT or other AI tools are prohibited in all classes, assignments and assessments at the University. If teachers suspect that students are using ChatGPT or other AI tools, they can ask students to discuss their homework, conduct additional oral tests and exams, etc.

Last week, Journal of Jinan (Philosophy and Social Sciences Edition) and Journal of Tianjin Normal University (Basic Education Edition) issued a statement stating that they would not accept any articles signed by large language model tools (such as ChatGPT) alone or jointly.

In the United States and Europe, more universities and academic journals have banned ChatGPT earlier.

However, simply proposing "blocking ChatGPT" can eliminate plagiarism and plagiarism?

Previously, the popular "AI face-changing" application has spawned a number of "anti-AI face-changing" identification tools. Now, is there any way to identify papers generated by ChatGPT?

Tang Jiayu, vice president of RealAI, which focuses on the identification technology of synthetic audio and video, told TechWeb that the difficulty in identifying ChatGPT products at present lies in the fact that there are few differences between texts written by machines and humans. The results of machine synthesis texts can better follow the rules of human writing from both structural and semantic points of view. At the same time, the way a real person speaks may also be changeable and structurally incorrect. Therefore, it is difficult to distinguish whether ChatGPT is generated from the text itself. At present, RealAI's ability to discriminate against synthetic text is already under construction, and a demo will be launched in the near future.

Digital watermarking for "anti-ChatGPT" is wishful thinking In the "anti-ChatGPT" technology stream, digital watermarking technology was once heated. But can digital watermarking really be used to anti-ChatGPT?

Digital watermarking is a technology that can be used to protect digital content such as text, images, audio and video from unauthorized copying and distribution. Digital watermarking can embed some hidden identification information into digital content, which usually does not affect the digital content itself, but can help identify the true source of digital content and copyright information.

People have two imaginations about anti-ChatGPT with digital watermarking.

One is to digitally watermark the content generated by ChatGPT to indicate that the content is written by ChatGPT, realizing anti-ChatGPT. But this requires the cooperation of AI content generation companies such as ChatGPT, that is, they are willing to do so.

ChatGPT developer Open AI has previously said it is considering ways to watermark content generated by AI systems to help identify whether text is AI. So far, however, Open AI has not taken this action.

Even if AI content generation companies are willing to add digital watermarks to the generated content, in the opinion of relevant industry experts, digital watermarking is only a technical means and cannot solve all plagiarism problems. Some people may use various methods to remove digital watermarks, or to circumvent detection of digital watermarks by changing the order of some words or sentences.

AI-generated content is usually generated based on a large amount of training data and model parameters. Another imagination of using digital watermark to counter ChatGPT is to mark the content with digital watermark to indicate the copyright identity and avoid being used for training by AI models such as ChatGPT, thus preventing imitation and plagiarism.

But this is a bit wishful thinking. Digital watermarking does not guarantee that content will not be used to train models. Some criminals may use various technical means to try to remove the digital watermark, or to bypass the detection of the digital watermark by modifying the article. Moreover, some unscrupulous people may ignore digital watermarks and steal copyrighted content as training data.

"Anti-ChatGPT" These tools can be used in combination with several tools currently used to detect whether articles are generated by AI, including:

OpenAI GPT-3 Playground: An online application developed by OpenAI to test and explore the capabilities of the artificial intelligence language model GPT-3. In this application, you can enter some text and the GPT-3 model automatically generates the next sentence or the complete article. This application can also be used to test whether an article is generated by a language model such as GPT-3.

Grover: A tool developed by the Allen Institute for Artificial Intelligence to detect fake news and fake articles. Grover can analyze the language style and structure of an article and try to distinguish between articles generated by humans and artificial intelligence. The tool can also identify common forgery techniques and tricks.

AI21 Writer's Studio: An online writing tool developed by AI21 Labs that provides users with automated suggestions and editing services to help them write more fluent and accurate articles. This tool can also be used to detect whether an article is generated by artificial intelligence.

Botometer: A Twitter bot Detection Tools developed jointly by Indiana University and the University of Southern California. Botometer can analyze the activity and behavior of a Twitter account to determine whether it is managed by a real user or an automated bot.

These tools can be used to detect whether an article or paper is generated by artificial intelligence, but it should be noted that they are not 100% accurate.

Therefore, when assessing whether an article or paper is written by AI, it is best to combine multiple methods and techniques for comprehensive analysis and judgment.

Taking ChatGPT as an example, Tang Jiayu mentioned to TechWeb that "identifying 'ChatGPT products' is to make good use of the duplication mechanism and technology to prevent ChatGPT from generating academic articles by rewriting, summarizing, etc.; Second, from the content itself, ChatGPT cannot fully guarantee the logical rigour and theme consistency of academic articles at present, which can be distinguished from the semantic level. "

In addition, from the case of undergraduate papers written by ChatGPT shared by UP Master B, check the source and background of the article. If the paper comes from an unknown or unreliable source, it needs to be evaluated more carefully.

As for the plagiarism phenomenon that ChatGPT may cause, Pan Xin, former COO of New Oriental Online, thinks that "this kind of worry is basically unnecessary." "" When there was no ChatGPT, there was no copying of papers and homework? You can believe one thing. If you can use technology to cause problems, you can definitely solve them by technical + administrative means. "

In the student party's view, ChatGPT is used to write papers. The data, arguments and even quoted documents are fabricated. It is suggested that ChatGPT should do some auxiliary work, such as giving suggestions on the outline or general direction of the paper.

Some students also think that self-discipline is needed."Anti-ChatGPT software has been launched one after another. The algorithm will only become more and more perfect. Even if it is now muddled, it is only a matter of time before it is detected. "

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.