news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

In the face of the accelerating AI wave, it is not realistic to ban AI across the board. We need to discuss it in different categories and situations to clarify the boundaries of AI creation, especially to distinguish between "AI-assisted" and "AI-led creation"

On the premise of defining the bottom line and boundaries, it is also worth observing and thinking whether the content and quality of AI-generated papers can replace or even surpass human originality.

"In the AI ​​era, education will change from teaching knowledge to teaching learning. Traditional knowledge transfer may become less important. It is particularly important to learn how to cooperate with AI and how to use AI to enhance one's own capabilities."

Text | Yu Xue and Liu Haoyu, reporters of Outlook News Weekly

During this year's graduation season, standardizing AI paper writing has become a hot topic on campus at universities such as Communication University of China and Tianjin University of Science and Technology.

Some universities have stipulated that the review of graduation theses (projects) must include detection of AI ghostwriting. Once the specified proportion is exceeded, the student will be warned and the evaluation will be cancelled. Some schools also list the reasonable use of AI, the re-examination method after exceeding the standard, and other detailed rules.

The attitude of colleges and universities towards AI writing papers is intriguing: on the one hand, with the rapid development of technologies such as generative artificial intelligence and machine learning, their functions almost cover the entire process of paper writing, such as retrieving information, collecting literature, processing data, drawing and tabulating, etc.; on the other hand, using AI to write papers and do graduation projects has also been questioned as to whether it constitutes academic misconduct.

A more profound thought is, with the coming of AI revolution, what changes will happen to the knowledge production model that was originally exclusive to humans? In the face of possible changes, how should the old knowledge inheritance model be adjusted to adapt?

Where is the limit of AI pen replacement?

By reviewing the relevant regulations currently issued by universities, it can be seen that each school mainly assesses the proportion of AI-generated content in the total volume of the paper, whether the usage is disclosed in detail, etc., to explore the boundaries and bottom line of using AI in the paper.

Duan Weiwen, director of the Philosophy of Science and Technology Research Office of the Institute of Philosophy of the Chinese Academy of Social Sciences, said that on the one hand, these initiatives show that universities are constantly exploring the boundaries and bottom lines of students' reasonable use of AI tools to assist scientific research. On the other hand, they also reflect that universities are striving to correct, honest and trustworthy scientific research attitudes and maintain scientific research ethics and academic purity.

At present, AI, as a technical tool, is becoming increasingly popular. According to a media survey, about 84% of college students in China have used AI tools.

Some students said that they use AI tools to collect data, quickly retrieve and organize information, translate data, draw and edit pictures, calculate and tabulate in their daily study and paper writing, which can greatly improve the efficiency of learning and writing. Some students also believe that AI search results are unreliable and the language is stiff, which is inspiring but of limited use.

Faced with the accelerating wave of AI, it is not realistic to ban AI across the board. We need to discuss it by category and situation to clarify the boundaries of AI creation, especially the distinction between "AI-assisted" and "AI-led creation."

In December 2023, the Ministry of Science and Technology issued the "Guidelines on Responsible Research Conduct (2023)", which clearly stated that generative artificial intelligence should not be used to directly generate application materials, generative artificial intelligence should not be listed as a co-achiever of the results, and unverified references generated by generative artificial intelligence should not be directly used, etc., clearly drawing specific boundaries.

In the view of Liang Zheng, deputy director of Tsinghua University's Institute of Artificial Intelligence for International Governance, an article mainly generated by AI is obviously not the author's original work and can be equated with "ghostwriting". If AI tools are partially used but the author has not verified it himself or concealed the use of them, it does not meet the requirements of scientific research integrity and is academic fraud.

Zhu Wei, deputy director of the Communication Law Research Center of China University of Political Science and Law, said that AI-generated content cannot be directly copied and pasted into papers for use without citation, nor can it be used as the main part of the paper. This is the key to drawing the boundaries of AI-written papers.

Experts remind that the corresponding regulations need to be adjusted to adapt to technological changes in the future, and further clarify the specific detection standards and punishment measures. On the basis of encouraging reasonable use and increasing the cost of violations, we should focus on cultivating students' scientific research capabilities and academic integrity education, and try to teach students to use AI tools correctly and maintain their pursuit of new knowledge and independent thinking ability.

At the 2024 World Artificial Intelligence Conference, a group of humanoid robots were displayed at the entrance of the exhibition (Photo taken on July 4, 2024) Photo by Fang Zhe/This magazine

Will AI's creativity surpass that of humans?

On the premise of setting the bottom line, it is worth observing and thinking about whether the content and quality of AI-generated papers can replace or even surpass human originality.

It is understood that AI writing tools are usually based on natural language processing technology and machine learning algorithms, which can understand and imitate human language expressions and generate logically coherent and fluent articles based on given topics and requirements. The core of AI writing tools is to learn and extract rules from a large amount of text data, but most of these contents are based on the combination and innovation of existing materials, which is far from the originality and innovation required by the paper.

Shen Yang, a professor at the School of Journalism at Tsinghua University, believes that AI’s ability lies mainly in combinatorial innovation, rather than fusion innovation, and it is even more impossible to achieve disruptive innovation. Even if the same AI tool is used by different entities, the role played by AI will be quite different.

This means that if the user has limited materials, AI tools can provide basic data collection, summarization and other services. If the user has made innovative breakthroughs in the field, the role of AI will be very limited.

Li Xin, deputy general manager and senior architect of Beijing Zhixiang Technology, reminded that although AI is not yet fully capable of creating highly professional and high-quality papers, with the speed of development of AI technology, its role in paper creation is likely to change from an auxiliary tool to decision-making participation.

As AI continues to test the boundaries of human capabilities, experts believe that instead of worrying about whether AI can replace human creativity, it is better to think about how humans can coexist with AI and how to highlight human value and amplify human capabilities by matching appropriate AI tools.

At the 2024 World Artificial Intelligence Conference held recently, Baidu Chairman and CEO Robin Li said that AI plays more of a "co-pilot" role in this wave, and ultimately it still needs people to control it. It is not a competitor to people. "AI only assists people in their work, not replaces them. It can make people work more efficiently and with better quality."

In addition, in the unknown areas that AI cannot explore, this part of the "initiative" still lies with humans. "In the AI ​​era, human insight and judgment are more valuable, and aesthetics, judgment, decision-making, inspiration, etc. need to be firmly grasped in human hands," Shen Yang said.

How will education change in the AI ​​era?

Although AI is positioned as "assistive" in its relationship with humans, it cannot be ignored that the strong support of AI is giving rise to tremendous changes in educational models and educational concepts, and traditional talent training models are experiencing impacts.

Many students said that in addition to tasks that must be completed independently, they have unknowingly formed the habit of "asking AI when in doubt", but they will only judge to what extent they can seek help from AI when using AI. "This will affect the ability to think independently to some extent."

Liu Yongmou, a professor at the School of Philosophy at Renmin University of China, said that the impact of large language models on education has just begun to emerge and will intensify in the future. "In the future, students may live in an environment assisted by artificial intelligence, so related products cannot be banned in schools and academia to prevent the students they train from lagging behind society."

Facing the impact of AI, people have different attitudes:

Zhang Miao, a "Haidian mom", has a child who is in the fourth grade of primary school in Haidian District, Beijing. The child has already completed the Chinese, math and English courses for the first year of junior high school. Zhang Miao plans to complete the high school courses before the third year of junior high school, and complete advanced math before the third year of high school, and get a GRE score of 325 or above. However, the amazing ability of AI big models in learning existing knowledge in the past two years has made her worry about whether the focus of future talent competition will change. Once it changes, will the plan she set for her child lose its competitiveness?

Unlike Zhang Miao, Ma Jing, whose child is in the fifth grade of Haidian District Primary School, is optimistic about the future. Ma Jing said that she has never constrained her child's energy to learning knowledge itself, but has constantly accompanied her child to explore various possibilities. Although her child's grades are not outstanding, Ma Jing believes that it is more important to explore the child's innate qualities than to master the knowledge of higher grades early. "I hope that the child will eventually find the track he is good at, and various technical tools including AI can help the child focus on this track and make up for his other shortcomings."

Liang Zheng believes that there is a certain deviation in society's understanding of education from cognition to practice, that is, education is equivalent to knowledge, and it is believed that knowledge means ability. "In the AI ​​era, education will change from teaching knowledge to teaching learning. Traditional knowledge transfer may become less important. It is particularly important to learn how to cooperate with AI and how to use AI to enhance one's own abilities. This actually puts higher demands on people's abilities."

In other words, if education in the past was about cultivating people who can answer questions correctly and give correct answers, then in the future AI era, students may not be required to answer questions correctly and give correct answers, but to ask questions and think of their own.

From this perspective, future education will gradually "converge" in terms of knowledge imparting, but the model will be more diverse in terms of stimulating individual potential and creativity.

Looking to the future, Chu Zhaohui, a researcher at the Chinese Academy of Educational Sciences, suggested that schools and parents teach students in accordance with their aptitude, focus on cultivating children's media literacy and technical literacy, and enhance their critical thinking. In his view, AI technology has changed the tools, content, and environment of education, but the core values, concepts, and basic laws of education will not change due to technological advances, and the specific logic and growth laws of human development will not change with changes in external forms.

The thinking mode of generative AI mainly relies on big data models, which is different from the thinking mode and cognitive mode that humans have evolved over millions of years. When humans face "others" with completely different skills and cognitive logic from their own, they cannot simply follow or imitate, but must maintain their uniqueness and focus on great wisdom that transcends specific knowledge levels.

Many experts suggest that in the AI ​​era, education should help build students' ability to ask questions, surpass AI's ability to solve problems, and see through the essence of things. "Only by thinking clearly about what problems you ultimately want to solve and what the ultimate purpose of learning is, can you better make AI your tool."

Every scientific and technological progress in history has inevitably promoted the transformation of knowledge production and education models. At present, we are once again standing at the crossroads of history. Facing the rapid development of science and technology, we should be in awe.

As the British writer Eliot said: "What good is opportunity to those who cannot take advantage of it?" The real challenge is never the technology itself, but how we can wisely use this unprecedented opportunity, neither getting lost in the torrent of technology nor giving up the brilliance of human wisdom, so that every technological advancement can become a step for human civilization to leap forward. (At the request of the interviewees, Zhang Miao and Ma Jing are pseudonyms) ■