The theme of this year's International Museum Day is "Museums for Education and Research", which not only clarifies the traditional function and role of museums as centers for the dissemination of knowledge, but also highlights the contemporary value, innovation and creative potential of museums in the education system. In the era of artificial intelligence, the three forms of existence of museums have changed dramatically, how can museums (plural) devote themselves to education and research?
International Museum Day is celebrated annually on 18 May, and this year's theme is "Museums for Education and Research". This theme is another profound emphasis and repositioning of the role of museums in modern society. It not only clarifies the traditional function and role of museums as centers for knowledge dissemination, but also emphasizes the contemporary value, innovation and creative potential of museums in the education system.
Generally speaking, there are three "forms" of existence of museums, namely, as cultural and educational institutions and organizations, as buildings and venues (places), and as cultural outputs. The "three E's" (education, enlightenment, and entertainment) of the basic functions of museums are all based on these three forms or states of existence of museums. In the era of artificial intelligence, these three forms of museums have still undergone tremendous changes, how can museums (plural) devote themselves to education and research?
Although the International Council of Museums (ICOM) does not mention AI in its elaboration on the theme of International Museum Day, this topic should be very important and urgent at this moment.
Shanghai Expo East Pavilion
Whether or not they agree with the "first year" of artificial intelligence, it is generally acknowledged that we are entering a "new era of unprecedented unknowns", the "evolution speed" of ChatGPT and OpenAI changes almost every week, and people of insight generally realize that the impact of disruptive technologies of artificial intelligence on learning methods and education cannot be overestimated in any way, because the way knowledge is generated and transmitted has changed, so this is a real revolution. A large language model (LLM) is an intelligent neural network that is capable of generating text that reflects human language by examining large amounts of text data. It can be seen as an advanced form of autocomplete, with surprisingly predictive abilities, a solid "understanding" of language, context, and meaning, and the ability to respond coherently and knowledgeably on different topics. Recently, the distribution and interactive display of various new products of OpenAI and GPT4o have allowed us to see the full advent of the era of artificial intelligence.
A case of 3D knowledge graph of large language model (courtesy of Dr. Huang Jun, Shanghai Advanced Research Institute, Chinese Academy of Sciences)
As non-formal educational institutions, museums have the flexibility that formal educational institutions do not have, and moderate foresight or even partial presumptuousness is a "mine-clearing" behavior, and should take the initiative to meet the challenges of new technologies. We have advocated that China's cultural and museum community should actively participate in the creation of the metaverse and grasp the strategic direction of the new round of scientific and technological revolution and new opportunities for industrial transformation. For museums, it is really dizzying to face the more than 100 large language models on the market, and there is always a sense of confusion about what to do. In the online world, there is "good news" about technological progress almost every day, and it is inevitable that museums have a phobia of technology choice. Learn and try to use these large language models, such as Chatgpt, Wenxin Yiyan, iFLYTEK Xinghuo, etc., and you will find that these mainstream language models cannot answer professional questions about museums, and even some of the answers have factual errors. This is obviously not a problem with language models, but rather that the research and development of language data in museums is a "more professional" job. Due to the huge amount of knowledge and high degree of professionalism in the field of cultural museums, such as the names of ancient bronzes have no connection with actual life, these knowledge needs to be provided by cultural and museum personnel to R&D personnel as data for training, and if only relying on general language models can not meet the accurate extraction and application of professional knowledge. When I communicated with Dr. Huang Jun at the Shanghai Advanced Research Institute of the Chinese Academy of Sciences, he repeatedly emphasized that in all "interactive" areas of museums, such as museum learning applications, it is important to combine natural language processing (NLP) related technologies and domain expertise, and one of the most important and urgent tasks is to train the language model of massive digitized collection data, research results and user data, so as to improve the quality and accuracy of the generated text. in order to ensure that the information and explanations provided are true, accurate and factual, as well as in line with the cultural values of the museum. In order for museums to actively participate in large language models, the professional value of museums must be valued.
At present, the large model commonly used on the market (Photo courtesy of Dr. Huang Jun, Shanghai Advanced Research Institute, Chinese Academy of Sciences)
Since the development of museum language models (or proprietary language models) is based on "external" open applications, the professionalism of museums has a certain "intrinsic" in multiple dimensions such as data, information and knowledge. Therefore, it is the first step of the museum language model to provide an intelligent tool for museum professionals to meet their needs for information inquiry, exhibition planning, cultural relics interpretation, and educational activities in their work. Some natural museums and art museums in Europe and the United States have begun to change the mode of manual explanation, and promote the "knowledge service in the background", after the front-end audience asks questions or interacts, a group of experts and docents sit on the terminal platform to "triage", and quickly transmit the answers to the front-end audience. This relatively time-consuming and laborious approach is now changing. By the time the museum's proprietary language model matured and was applied, the back-end consulting team had been replaced by the system to develop newer content. The Shanghai Advanced Research Institute of the Chinese Academy of Sciences and some "advanced" museums in China, such as the Shanghai Museum and the Jingdezhen Imperial Kiln Museum, have been cooperating on museum-specific language models, and have completed the first phase of the results, such as the Chinese ceramic gene bank has played an effective role in exhibitions and education. Due to the computing power of AI technology, which is always open, outgoing, and self-renewing, these proprietary language models can easily be used in education and research, and then embedded in exhibitions, education, events, guided tours, and various derivative activities.
Museum language model (courtesy of Dr. Huang Jun, Shanghai Advanced Research Institute, Chinese Academy of Sciences)
Museum-specific language models are not "fashionable" and "brain-hot", but how can I make sure that they are used and effective? Planning and design is very critical, in the initial functional module, must be added collections, exhibitions, education, research, audiences and other settings, some cultural and creative museums should also take this part of the data as the basic module, which is ignored in the previous development planning or special planning. The cultural and creative purchase data in the audience data is a very important factual data in the audience behavior. It is best to "compile data" such as visitor surveys and interviews over the years in the museum. Considering that the average annual number of visitors in China has exceeded 1 billion, the time is ripe for the establishment of a regional or national audience database, which is a supreme credit for the language model training of museums around the world.
The language model can be applied to some work of the museum to provide guidance and suggestions for exhibition planning, including the selection of exhibition themes, the layout and display of exhibits, the writing of exhibition outlines, the interpretation of exhibits and the writing of labels, etc., to help staff design wonderful exhibition content. Language models can provide suggestions for the design and planning of educational activities, provide appropriate educational content and interaction methods for young people of different ages, and help educational event planners determine the content of lectures and practical activities related to the theme of the exhibition. The language model can provide information on academic research results, academic literature and professional journals in the field of culture and museums, and provide support for the academic research of library personnel and collection data, which greatly improves the efficiency and accuracy of research data acquisition. In the interpretation and knowledge interaction of exhibits, in online and offline exhibitions or guides, language models can provide intelligent explanation, intelligent Q&A, and knowledge interaction for the audience. Provide detailed interpretation and background knowledge of specific exhibits to help better understand the historical, cultural and artistic value of the collection.
Museum language model (courtesy of Dr. Huang Jun, Shanghai Advanced Research Institute, Chinese Academy of Sciences)
The language-specific model of museums will certainly promote the transformation of museum institutions, first, the significant improvement of work efficiency. The significance of the museum language model is to improve the efficiency and quality of the internal work of the museum. It can provide immediate information and resources to staff, avoiding duplication of effort and waste of resources. For example, when planning an exhibition, staff can use language models to quickly obtain relevant information to better design the content of the exhibition and interpret the artifacts. Second, the degree of specialization of museums and the overall literacy of the industry have been significantly improved. As a learning tool, language models can help museum staff continuously improve their professional knowledge and skills, and promote the innovation and development of museums. Through the dialogue with the language model, the library staff can obtain professional advice, refer to academic literature, and learn about the latest research results, so as to improve their professional quality. Third, the level of audience integration and the depth of interactive experience have been significantly improved. The functions of intelligent explanation, intelligent question and answer, and knowledge interaction provided by the language model for the audience have enhanced the audience's viewing experience, enhanced the knowledge connotation of the exhibits, and promoted the museum's knowledge dissemination ability and public service level. In short, the three forms and states of existence of museums will also gradually change, and the revolution of museums is coming.
Subo exhibition site
(The author is a professor of museology, Shanghai University)