The full text of the ＂Research Report on the Development Strategy of Cultural and Technological Integration in China＂ was released.
On September 21st, 2023, it was officially released in the main forum of China (Nanjing) Exhibition and Trade Fair on Cultural and Technological Integration Achievements. The Research Report on the Development Strategy of Cultural and Technological Integration in China (2023) (hereinafter referred to as the Report) was compiled by the Institute of Creative Industry Technology of China Renmin University.
The Report attempts to build a circle map of "one core and three circles" science and technology supporting the construction of modern civilization of the Chinese nation in theory, and further focuses on the annual hotspot in the field of culture and science and technology-generative artificial intelligence, deeply analyzes its development process, technological singularity, industrial practice and scene application, and looks forward to the eight focuses of cultural and technological integration in the future AIGC era, namely, data credibility, model opportunity, substitution anxiety, copyright dilemma, content mining, value derivation and cultural governance.
Research Report on the Development Strategy of Cultural and Technological Integration in China (2023)
First, the scientific and technological support route of the modern civilization of the Chinese nation
(A) science and technology to support the construction of modern civilization of the Chinese nation "one core and three circles" technical route
The deep integration of culture and science and technology plays a cohesive role in promoting the innovative development of the excellent cultural genes of the Chinese nation and the new generation of digital technology, and promoting the construction of modern civilization of the Chinese nation.
Chinese excellent traditional culture, advanced socialist culture and revolutionary culture, as the crystallization and essence of Chinese civilization, are the roots and souls of the Chinese nation and constitute the cultural core of modern Chinese civilization. As the primary productive force, science and technology are gradually embedded in cultural inheritance and development, deductive presentation, exchange and communication, etc., extending the length of cultural continuity, exploring the depth of cultural presentation, expanding the breadth of cultural communication, and building a "one core and three circles" technical route that science and technology support the construction of modern civilization of the Chinese nation.
The "One Core and Three Circles" Technical Route of Science and Technology Supporting the Construction of Modern Civilization of the Chinese Nation
1. Technology extends the length of culture.
-The integrated development of culture and science and technology promotes the continuous inheritance of Chinese civilization. First, science and technology have extended the space for innovation and development of culture in the digital age, given Chinese excellent traditional culture new life and new interest, promoted the transformation of cultural resources into digital forms, realized the living inheritance and digital sharing of culture, enhanced the sustainability of culture, and promoted Chinese civilization to achieve "digital immortality". Second, science and technology help to trace the origin of civilization, and through the dialogue with history through the power of science and technology, reveal the origin, formation and development of Chinese civilization, and contribute to reappearing the brilliant achievements of Chinese civilization. Third, science and technology empower the culture of live transmission. The development and application of next-generation information technologies such as big data, cloud computing and artificial intelligence expand the existing forms of traditional culture. Data centers such as cultural gene bank, database and resource bank can permanently preserve and live Chinese culture, promote the integration and close connection of cultures, promote the panoramic presentation of Chinese culture, and share the digital achievements of Chinese civilization with the whole people.
2. Science and technology explore the depth of cultural presentation
-Science and technology have revolutionized the way of narrative and deductive presentation of cultural stories, matching the cultural needs of "online migration" in the new era. On the one hand, technology has innovated the spatial expression of cultural resources and broken through the one-way interaction mode between traditional culture and the public. Specifically, the presentation mode of cultural resources is changing from "single line and offline" to "online presence", and cloud exhibition, digital museum, holographic phantom imaging display and other modes have emerged. On the other hand, technology has realized the two-way interaction between Chinese civilization and the public, and the exhibition form characterized by immersive and interactive experience has revolutionized the past "unilateral and thin content" cultural presentation posture. The market size of China’s immersive industry is expected to reach 52 billion yuan in 2022, and it is expected to exceed 250 billion yuan in 2026, and the compound growth rate will reach 48.1% during 2022-2026 (source: China Immersive Industry Research Report 2022). Immersive format is becoming a new track of cultural industry. Virtual reality technology has greatly improved the expressive force of culture, and artificial intelligence technology has promoted the interactive development of cultural experience, and more perceptive cultural digital scenes such as digital cultural relics and digital historical and cultural blocks have emerged, which has become a new model for the charm of Chinese civilization.
3. Science and technology expand the breadth of cultural communication
-Science and technology have promoted the exchange and spread of Chinese civilization. First, technology has further expanded the boundaries of cultural exchanges. Under the new technical support system, digital media such as VR, 5G, AR and cross-media platforms have emerged, realizing the organic integration and complementary advantages of digital media and traditional media, opening up a long-distance, high-precision and low-cost communication path for Chinese excellent traditional culture, broadening the breadth and depth of communication and communication of Chinese civilization, and effectively promoting the flow and integration of Chinese civilization. Second, technology has further enriched the forms of cultural communication, breaking through the centralized mode of communication in which graphics, text, audio and video are the main parts and there is little interaction and communication. Driven by digital technology, digital people, virtual anchors, virtual studios, non-live broadcast rooms and other lightweight and interactive new forms of communication have emerged, resulting in efficient and intelligent digital media tools, which show the unique charm of Chinese civilization in an all-round and three-dimensional way. Third, technology has further released the power of cultural communication, creating opportunities for each individual to freely accept and participate in cultural expression. The popularity of smart terminals such as mobile phones has lowered the threshold for content manufacturing and content dissemination, greatly stimulated the public’s willingness to express culture and creativity, and the arrival of "national media" has accelerated the communication efficiency of Chinese civilization.
Second, the culture and technology in 2023-the moment to break through the technological singularity
In November 2022, artificial intelligence represented by ChatGPT swept the world, becoming the fastest-growing application in history, which was called "artificial intelligence milestone". Phenomenal AI application has set off a wave of technological revolution, and promoted the integration of culture and technology into the process of intelligent, digital and interconnected development. Technological innovation has become the key variable leading the transformation of the cultural industry and the biggest increment to achieve high-quality development.
(A) technical singularity: the key technology is about to detonate digital efficiency.
Throughout the development process of cultural science and technology industry, the intervention of key technologies is the key to industrial transformation and upgrading. In the era of switching from Internet to mobile internet, technologies such as database, server and web design have opened up information transmission channels, and the information exchange paradigm has been transformed by portal websites. In the era of switching from Internet to mobile internet, technologies such as 4G intelligent terminals have driven the explosive presentation of digital content industry with new media such as short video, and the industrial scale has grown exponentially. However, the conversion from mobile internet to Internet of Everything is limited by key core technologies, and the concept is difficult to land. There is an efficiency dilemma in the overall planning of big data resources, and the product is difficult to form due to the limitation of computing power. It is urgent to use key technologies to detonate the efficiency of industrial development. As a new technological singularity, AIGC is expected to become the link point for the smooth transition from mobile internet to Internet of Everything, reconstruct the thinking mode, production process and service system, open up new business opportunities and business models, and usher in the "iPhone moment" of the cultural industry.
Break through the technological singularity moment
(B) qualitative change era: the wave of cultural and technological integration under the outbreak of AIGC
Under the wave of AIGC, the scientific and technological attributes of the content industry are increasingly enhanced, and the demand for personalization, fragmentation and customization drives scientific and technological innovation in the opposite direction. The AIGC model based on the content industry began to fluctuate and evolve, exploring the development path from long-chain dialogue to "literary and graphic frequency".
The first stage is the exploration stage of deep learning model. The data and elements of culture are the key of AIGC model, and the deep self-learning model provides a key technical base for the chain of "integration-deconstruction-regeneration" of cultural data. In 2014, the Generative Adversarial Networks (GAN) based on the cooperative zero-sum game framework appeared, which became the exploration cornerstone of the diffuse development of deep learning models. In 2017, Transformer architecture relied on two components, namely encoder and decoder, which brought a subversive breakthrough to the field of natural language processing and laid a model foundation for the deconstruction and regeneration of content.
The second stage is the construction stage of the basic structure of "Wen Sheng Wen". At this stage, the self-deep learning algorithm and model of AIGC have gradually matured, and the generative AIGC has gradually shifted from theoretical concepts to practical applications, and the basic framework of "Wen Sheng Wen" has gradually taken shape. In 2018, Google released the Bert (Bidirectional Encoder Presentations from Transformers) model, and social networks, recommendation systems and other fields closely related to the cultural industry entered a critical period of research. In 2019, GPT-3 has been able to complete high-quality tasks such as news and stories; Baidu launched ERNIE artificial intelligence system, which greatly improved the ability of language understanding and generation. Natural language processing and machine image formation have entered a new stage of development, and AIGC has gradually penetrated into cultural core fields such as publishing industry and journalism.
The third stage is the multi-modal integration stage of "Wensheng Graphic Frequency". The technology of "Wen Sheng Tu" is becoming more and more mature, and some AI creation tools have been able to realize efficient and high-speed "text input-image generation" conversion. With the characteristics of automation, high precision and customization, it has become the connection between text and image modal conversion, and AIGC has entered the multi-modal development stage of "Wen Sheng Tu Frequency". As a new track, the limitations of the "Wensheng Audio-Video" model still need to be broken, with simple logic, content reorganization and other structured content as the main content, and the generation scale does not match the generation quality, so it is difficult to meet the creative needs.
Cross-module and multi-modal integration stage of "Wensheng Graphic Frequency"
(C) Hundred Models Wars: the gap, situation and structure of AIGC big models at home and abroad
The emergence of ChatGPT has set off a wave of generative artificial intelligence. According to IDC statistics, the global artificial intelligence market will reach 82 billion US dollars in 2026, and AIGC at home and abroad presents a development trend of "hundred-mode war". The resource endowments such as data volume and data center have become an important symbol of AIGC’s soft power.
China has a large amount of data, but there is a big gap in data carriers. The complex deconstruction and iterative updating of AIGC model requires a huge training library as the bottom support. According to IDC data, the data volume of China will increase from 23.88ZB in 2022 to 76.6ZB in 2027, with an average annual growth rate of CAGR of 26.3%, ranking first in the world. Data center is an important spatial carrier of data informatization and a key platform support for big data storage and system operation. According to Synergy Research data, large-scale data centers in the United States account for 49% of the world, ranking first, while China ranks second with 15%, and there is still a big gap.
The number of patents and talents is unbalanced. AI patent is the key to break through the pain point of core technology in China. According to the statistics of LexisNexis PatentSight, as of 2021, in the TOP10 list of global AI patents, four Chinese enterprises are on the list, among which Tencent Group ranks first with 9,614 patents, and Baidu ranks second with 9,514 patents, which has strong patent innovation strength. However, according to the list of the most influential AI scholars in the world in 2022-AI 2000, the number of high-impact AI scholars selected in China is only 232, and the imbalance between the number of high-quality AI talents and the number of patents is obvious.
The foundation of the model is weak, and some of them are still based on foreign models. Foreign AIGC model research started early and has a strong foundation, and Transformer and BERT have become the bases for the development of many large models. Limited by the rigid requirements of high cost, large computing power and strong technology, the independent research and development scale of large models in China is small, and some AIGC models are still based on foreign research and development models.
"Follow-up" big model field, deeply cultivate small model vertical track. In the field of large models, foreign large models such as ChatGPT "lead" the industrialization and large-scale development, and the training accuracy and model accuracy are good. China’s large-scale model is still in the "follow-up" stage, which has a certain advantage of being a latecomer, but it still needs to break through the critical point of scale benefit and release greater efficiency. In the field of professional models, foreign countries focus on the innovation of visual expression and auditory paradigm of AIGC, and Stable Diffusion is representative in the field of AIGC image generation. China’s models exert more efforts, emphasizing wide range and multi-layout, deeply cultivating games, movies, cultural tours and other tracks, and actively exploring new tracks for AIGC+ industrial transformation.
Comparison of large AIGC models at home and abroad
(D) Scenario application: Multi-link infiltration of digital intelligence integration
On the premise of breakthrough development of AIGC’s deep learning and computing algorithm, AIGC’s content creation paradigm and service model realize self-iteration, and gradually deeply empower and penetrate into the creative chain of cultural content from production to review, resulting in five scenarios, such as content creation and cultural security.
AIGC empowers the creation and generation of universal cultural assets. With the support of natural language processing technology and analysis algorithm, AIGC data can be labeled and transformed into elements, which has certain reusability, extensibility and scene adaptability, and serves multi-modal structured scenes such as game NPC dialogues and article outlines.
AIGC empowers the auxiliary production of professional cultural assets. The production content of professional vertical track has the characteristics of personalization, customization and fragmentation. AIGC can provide strong professional content materials, assist the creative behavior of artificial content with high creativity and precision, and improve the creative efficiency of human producers.
AIGC empowers cultural tourism consumption decision-making and personalized service. Large-scale data volume and strong computing power will guarantee the decision-making ability of AIGC’s intelligent cultural tourism. By analyzing the consumption history of cultural tourism and the behavior preferences of cultural tourism destinations, the model can automatically analyze the personalized and fragmented preferences of cultural tourism, improve the decision-making accuracy of customized behaviors of "thousands of people", and empower self-service customer service in scenic spots, intelligent scene navigation and intelligent cultural and creative marketing.
AIGC enables content security and auxiliary audit. The blowout development of digital content brings great pressure to content inspection, review and screening. Through targeted pre-training, AIGC can assist in the review of words, pictures and videos, and achieve cost reduction and efficiency improvement in the content review process through human-computer collaboration.
AIGC empowers the structure and knowledge of content information. With the support of artificial intelligence analysis algorithm and natural language model, AIGC dismantles the content elements through the "deconstruction-classification-reconstruction" link of cultural content, realizing the structure, knowledge and systematization of cultural content and serving the cultural heritage knowledge map and other scenes.
Scene Application: Multi-link Infiltration of Digital Intelligence Integration
(E) Core issues: AIGC front-end specifications and link walls
The rapid development of AIGC has brought great changes in the content paradigm and business model for the digital culture industry, but the imperfect regulatory framework of existing laws and regulations and model training technology have brought great challenges to AIGC compliance and iteration.
The nonstandard front-end training leads to the "incorrect beam" of the model. On the one hand, the data is not standardized and the credibility is low. Corpus is the key foundation of AIGC model training, but the high-quality Chinese corpus in China is small in scale and difficult to obtain, and the problem of "data island" is serious, which urgently needs high-quality general monolingual corpus to support model training; Internet content data is large in scale and difficult to obtain, which can provide data scale guarantee, but the quality is difficult to control. On the other hand, the model training frequency is low and the iteration is weak. Influenced by model complexity, computational power basis and other factors, the model training frequency is difficult to reach the ideal level, and the low frequency is difficult to meet the needs of high precision training. The pre-training level of large-scale models is unstable, which is prone to problems such as training interruption, uneven language understanding function, and the effect of data interactive training is affected.
The key link of AIGC is "building high walls". On the one hand, the storage dilemma causes the memory bottleneck. At present, AIGC memory is difficult to provide strong support for the scale data of complex models, and the storage efficiency is low. It takes about 10 days to preprocess the 100 TB data. The storage speed is slow, and the limited data memory can not guarantee the file loading speed, so it is difficult to meet the model training requirements. On the other hand, AI chips build technical barriers. AI chip is an important guarantee for AIGC model operation and a technical support for model training efficiency. At present, the production technology of AI universal chips is demanding, and there is a lag in mass production, and technical barriers are still difficult to break in the short term.
It is difficult to link the real economy because the back-end of the model is weak. In terms of scene application, AIGC has penetrated deeply into the fields of e-commerce, publishing, art and so on, relying on the interactive and dynamic characteristics. However, due to factors such as poor demand docking and difficult guarantee of supply quality, it is difficult to perfectly match the application performance of AIGC with the scene requirements; In terms of charging mode, the application of AIGC in China is still in the "price war" stage of free drainage and low-cost entry, and it is difficult to realize the conversion from "large flow" to "high value" in a short time, and the link path to the real economy is still unclear.
Core issues: AIGC front-end specifications and link walls
Iii. Eight Focuses of Cultural and Technological Integration in the AIGC Era
Based on the above analysis of the development history, scene application and core issues of AIGC, this Report predicts and puts forward eight focuses of cultural and technological integration in the AIGC era, and explores the key links and current trends of AIGC in the digital content creation ecology.
Focus 1: data credibility-the data specification in the upper, middle and lower reaches of ——AIGC is the primary issue.
1. Unreliable and nonstandard or become two stumbling blocks of AIGC
At present, the AIGC field has basically formed a data content production chain of "collection-storage-operation-deconstruction-reassembly", but some links are still not perfect, and there is a "gray zone" in industry supervision.
The corpus is small in scale and low in quality, so the training data is facing the challenge of "credibility". As a key carrier of non-data structure, corpus plays an important role in model training. Generally speaking, Chinese corpus faces the following three "credible" challenges. First, the corpus content is small. According to the data, by the end of April 2023, the number of mobile Internet users in China reached 1.485 billion, but the Chinese content of the Internet only accounted for 1.5% of the global Internet content, and there were few available training data. Second, it is difficult to obtain corpora, and some high-quality corpora such as business, scientific research and finance still have "data barriers" and are difficult to access publicly; Third, the corpus is coarse in granularity, which is influenced by the method, efficiency and cost of corpus tagging. Some corpora have large granularity of semantic tagging, so it is difficult to extract effective information from the chain of "tagging-retrieval-mining-analysis-training".
Improper model training brings "AI illusion" and causes credibility crisis. The increasing complexity of Large Language Model (LLM) and the massive influx of data lead to "AI illusion", that is, AIGC generates self-confident responses whose contents are inconsistent with training data and objective facts. On the one hand, factors such as insufficient model data, incomplete data format and biased data labels have accelerated the emergence of "AI illusion"; On the other hand, the fitting degree of the model framework is low, and the training level of the model is difficult to meet the training needs of complex models, which makes it easier for high-level and multi-parameter models to produce error information.
Data credibility-the data specification in the middle and lower reaches of ——AIGC is the primary issue.
2. Three keys to the credibility of 2.AIGC’s upper, middle and lower reaches.
Under the premise of AIGC’s cross-disciplinary, cross-regional and cross-modal system complexity, we should focus on the whole industry chain of content production from data integration to model development, from tool platform to content creation, from industry service to market circulation, so as to ensure data standardization, industry compliance and product credibility with inclusive and prudent regulatory policies.
Upstream: guarding front-end data specifications by industry, classification and source. The data supplier is the key person in charge of the authenticity, transparency and standardization of AIGC content, and the compliance audit management of AIGC model pre-training data should be strengthened. First, the data security threshold should be defined by industry to avoid "one size fits all" management of data content; The second is to build a data security classification framework to ensure that data sources are legal and standardized; The third is to improve the granularity of corpus labeling by source and ensure the accuracy of data information extraction.
Midstream: Encourage the generation of protection and break the dilemma of false content with "technical empowerment". Ensure the creation and output of AIGC content from the aspects of model training and content production. First, increase the research and development of targeted technologies in the direction of model training and fitting degree, ensure the authenticity of content with "big scientific research" and break the dilemma of false content generation; The second is to encourage AIGC to create content by using differential privacy and other technologies to enhance the ability to protect personal privacy, trade secrets and other content, strengthen data protection and enhance the credibility of content.
Downstream: improve the risk classification system, and inclusive management and strict supervision coexist. Under the background of the large-scale and industrialized development of AIGC content, in order to further improve the ecology of digital content industry, it is suggested to adopt an inclusive and prudent three-level risk management system for the content generated by AIGC. For low-risk content, the model automatically judges and informs users of risks, emphasizing autonomy; In view of high-risk content, it adopts double insurance of manual audit and expert compliance assessment to pursue objectivity; In view of the unacceptable content, under the framework of government supervision, the content is removed from the shelves in a full model and scene, and the boundary is clear.
Data credibility-the data specification in the middle and lower reaches of ——AIGC is the primary issue.
Focus 2: model opportunity-Matthew effect of large model, and professional small model may be more likely.
Large-scale models in AIGC generally refer to machine learning models based on huge parameter scale and high complexity, which have the characteristics of large scale, strong analysis and high accuracy. However, due to the restrictions of high cost and high technology threshold, it is difficult for small and medium-sized enterprises to join the field and share the technology dividend. Therefore, the bottom logic of the large model should be parallel to the professional logic of the small model, and the professional small model with multiple scenarios will become the focus of AIGC marketization.
1. The big model is the cornerstone and base of development, but only a few of them have high threshold.
At present, generative artificial intelligence is deeply empowering the digital development of the cultural industry. The universal big model based on huge public data has become the base of the development of culture and science and technology. However, the characteristics of the big model, such as high training threshold and high cost barriers, are doomed to be that only a few market players can concentrate resources, give full play to their scale advantages and lead the research and development of "training big models". Small and medium-sized enterprises should try their best to avoid the "big training model" of scattered training resources and reduce the industry bubble and resource waste.
First, the training cost is high. The superposition of data, equipment and talents leads to the high training cost of AIGC. According to OneFlow, the maximum training cost of some large-scale LLM models can reach 12 million dollars. Second, the demand for chips is large. AI chips are an important execution unit for model training. The increase of AIGC market scale promotes the rapid growth of GPU demand. According to statistics, the daily demand for AIGC sites with about 250 million consultations is 30,382 chips, which makes it difficult for small and medium-sized enterprises to afford the huge chip scale. Third, the initial computational power is high, and the computational complexity required by the more and more complex AIGC model increases exponentially. The consultation volume of tens of millions of users needs to match 3798 advanced servers, and the cost is about 759 million US dollars. Matching high computational power in a short time brings great training pressure to enterprises.
2. Focus on small model professional logic, break through resource constraints and seek more opportunities.
At present, the high threshold situation of large-scale model admission cannot be changed in the short term, but the technology spillover benefits of large-scale model can also bring development opportunities to small and medium-sized enterprises, and help small professional models to continuously and deeply promote the scene expansion and development by means of transfer learning and resource sharing. The integration of the two will effectively release the long tail effect of AIGC business scenes and drive small and medium-sized enterprises to enter the market to share the development dividend.
On the one hand, strengthen the industry landing with "transfer learning". In the paradigm of "small model+transfer learning", by fine-tuning the training data of large model, small model can form new cognition and expand new scenarios at low cost, and accelerate the application of AIGC.
On the other hand, the production energy is activated by "light resources". Professional model is based on the re-debugging and retraining of general model. With the help of the existing platform of large model, professional small model can form a "symbiotic effect" with large model, providing lightweight solutions for small and medium-sized enterprises, realizing the professional path breakthrough of small model, and directly docking and meeting the market demand.
Model opportunity-Matthew effect of large model, professional small model may be more likely.
Focus 3: Substitution anxiety-Artificial intelligence is still a complement to human intelligence in the short term.
With the vigorous development of artificial intelligence, whether AIGC will replace human beings in the field of content creation has aroused widespread concern. How to treat the relationship between AIGC and human beings has become an important topic.
The creative paths of man and machine are quite different. AIGC’s creative path is mainly to reorganize the existing materials through algorithms to form the final works, which is to achieve content prosperity within the inherent boundaries; The creative path of human beings is based on the ideological perception of the world, expressing freely through various forms and forming the final works. This kind of creation provides more possibilities for exploring unknown boundaries and innovating content paradigms.
Substitution anxiety-artificial intelligence is still a complement to human intelligence in the short term.
The difference of creative path leads to the difference of man-machine efficiency. Based on the different creative paths, we can see that artificial intelligence and human intelligence have their own advantages in efficiency. On the one hand, the biggest advantage of AIGC is that it can efficiently handle structured tasks with complex calculations, which is suitable for the generation of universal cultural assets, with high efficiency and fixed boundaries; On the other hand, the greatest advantage of human beings is that they have the ability to perceive, learn, understand and communicate, and can handle all kinds of complex tasks. They are suitable for reviewing and revising general cultural assets and generating professional cultural assets, and are characterized by low efficiency and unlimited boundaries.
It is the difference between man and machine in creative efficiency that makes artificial intelligence unable to replace human beings in the short term. Only by complementary cooperation and breaking through the efficiency bottleneck and content boundary can they give full play to their comparative advantages and achieve maximum effectiveness.
Substitution anxiety-artificial intelligence is still a complement to human intelligence in the short term.
Focus 4: Copyright Dilemma —— Three-layer logic verification of ——AIGC copyright determination
Since AIGC came out, its copyright issue has aroused heated discussion from all walks of life. There are three main views on the copyright ownership of AI-generated works: one is the affirmative view that works generated by artificial intelligence are copyrightable as long as they meet the requirements of originality. For example, DABUS (artificial intelligence machine created by Dr. Taylor in the United States) has obtained patent authorization in South Africa; The second is conditional affirmation, that is, in content creation, human beings put their intellectual activities into it and then complete it with the help of artificial intelligence. This situation can be protected by copyright. For example, the case of Zarya of the Dawn in the United States narrowed the scope of copyright registration, only covering the "selection, coordination and arrangement of works created by the author and generated by artificial intelligence" formed by the author when writing this book, while those images automatically generated by Midjourney are not protected; Third, whether it is a qualitative point of view, that only people’s creative achievements can be protected by copyright as works, and the content generated by artificial intelligence is not protected by copyright. For example, the case of "A recent entry to paradise" in the United States determined that its works do not constitute copyright works. It can be seen that the copyright ownership of AI-generated works is still inconclusive in theory, and the judgments are different in practice.
Copyright Dilemma —— Three-layer Logical Verification of ——AIGC Copyright Confirmation
This Report holds that AIGC has the possibility of obtaining copyright on the basis of conditional affirmation, which needs to meet the following three conditions: First, whether the instructions given by human beings to machines are closed, that is, the input instructions should be original and closed, rather than using the existing instruction set; The second is whether to modify the content generated by the machine, that is, whether to further adjust and modify the generated content after inputting the instruction; Third, human beings have the creative control of the final results, that is, human beings have the ultimate control over the way of production and interpretation of works.
Copyright Dilemma —— Three-layer Logical Verification of ——AIGC Copyright Confirmation
Focus 5: Content Mining-AIGC realizes the transition from "false intelligence" to "true intelligence"
As a major breakthrough of artificial intelligence algorithm, AIGC can widely empower content fields from natural language processing, multimodal interaction, digital twinning and other business levels, and realize the transition from "false intelligence" to "true intelligence".
The differences between "false intelligence" and "true intelligence" are as follows: in content mining, the former is static and the latter is dynamically generated; In terms of content display, the former is mainly linear display, while the latter realizes deep interactive display; In terms of content demand, the former mainly meets the single demand, while the latter meets the multi-experience demand; In terms of content services, the former is a standardized service, while the latter realizes a customized service.
Content Mining-AIGC realizes the transition from "false intelligence" to "true intelligence"
AIGC has helped the intelligence of all important links in the content field. In content mining, AIGC can discover hidden cultural phenomena and laws through the study and analysis of large-scale corpus; In content management, AIGC realizes intelligent management of content through automation and intelligent technology; In terms of content display, AIGC realizes various forms of content display such as words, pictures and videos through multimodal interaction technology; In content interaction, AIGC model can automatically filter a large amount of information according to user needs and generate targeted content for real-time in-depth interaction; In content service, AIGC can automatically recommend content related to users’ interests by analyzing users’ historical behaviors and preferences, and realize personalized service.
Content Mining-AIGC realizes the transition from "false intelligence" to "true intelligence"
Focus 6: Value Derivation-AIGC promotes the transformation from service value to emotional value.
The application of AIGC in the field of cultural services promotes the transformation from service value to emotional value. Generally speaking, traditional cultural services are oriented by function value, with service function as the main body; AIGC intelligent service integrates personalized and customized services, providing emotional value in addition to functional value.
Traditional cultural services are guided by functional values. In terms of service acceptance, traditional cultural services are mainly single channel, and each channel is relatively closed; In terms of service efficiency, traditional cultural services mainly respond during working hours, and the efficiency is uncertain; In terms of service process, traditional cultural services generally have a relatively fixed process, characterized by standardization and linearization; In terms of service value, traditional cultural services mainly provide functional value.
AIGC intelligent service is guided by emotional value. In terms of service acceptance, AIGC intelligent services are accessed through multiple channels, and each channel can communicate with each other; In terms of service efficiency, AIGC intelligent service can realize all-weather response without distinction, with high efficiency; In terms of service process, the process of AIGC intelligent service is customized in real time according to the demand, which is characterized by personalization and diversification; In terms of service value, AIGC intelligent service provides interactive emotional value in addition to functional value.
Based on the above analysis, AIGC provides thousands of personalized services with low cost and high efficiency, which is more suitable for customers’ emotional resonance points and forms effective emotional value.
Value Derivation ——AIGC promotes the transformation from service value to emotional value
At present, some virtual bloggers and virtual anchors have certain ability of emotional dialogue and multi-modal interaction, which can provide some customized and personalized emotional value for immersive companionship. In the future, with the continuous improvement of technology, the reduction of computing power cost and the expansion of application scenarios, AIGC intelligent services with stronger emotional interaction function and better immersive companionship will have a broader development space.
Value Derivation ——AIGC promotes the transformation from service value to emotional value
Focus seven: cultural governance-AIGC brings new challenges to content security
With the emergence of ChatGPT, large language models of many domestic Internet companies have been launched one after another, and cultural governance is also facing new challenges of content security. First, AIGC can generate a lot of guiding content in a short time, which makes the competition pressure of public opinion game increase; Secondly, the multimodal interaction of AIGC makes the forms and types of illegal content complex and diverse, which makes it difficult to identify bad information after review; Thirdly, AIGC has formed a high professional barrier to the reproduction of compound professional knowledge, which leads to an increase in the threshold for judging the authenticity of content. Based on these factors, the content governance of AIGC needs to pay attention to key technical links such as data traceability, model algorithm review, and machine filtering review.
Cultural Governance —— The new challenge of content security brought by ——AIGC
Focus 8: Supervision Concept —— Considering the moderate balance between innovation and development and risk avoidance.
The free development of AIGC model will inevitably bring various risks such as data security, content security and ethical impact. However, if AIGC is fully and strictly supervised, it will restrict its model growth, technological innovation and universal potential. Therefore, the supervision of AIGC needs to consider the appropriate balance between innovation and risk aversion.
At present, the great changes brought by AIGC model challenge the traditional supervision. First of all, its application scope has expanded from subdivided fields to general scenarios, resulting in more decentralized supervision objects; Secondly, the uncontrollable development of technology (such as the emergence of large models) makes the risk uncertainty bigger, which makes it difficult for prior supervision to predict the risk; Finally, the technical iteration cycle is getting shorter and shorter, which leads to the problem of regulatory limitation.
The concept of supervision-considering the appropriate balance between innovation and development and risk avoidance
Based on the above considerations, AIGC needs an inclusive and prudent regulatory concept, and relevant departments in China have also carried out practical exploration in this regard. With the attention of all walks of life, the world’s first regulatory regulation in the field of AIGC came into being-the Interim Measures for the Management of Generative Artificial Intelligence Services came into effect on August 15, 2023. The Interim Measures stipulates that the state adheres to the principle of paying equal attention to development and safety, promoting innovation and governing according to law, takes effective measures to encourage the innovative development of generative artificial intelligence, and implements inclusive and prudent supervision of generative artificial intelligence services. The promulgation of this regulation has established the basic principles for the supervision of AIGC, but with the progress of technology, the perfection and implementation of AIGC supervision regulations need further exploration.
Culture and science and technology, like two wings of civilization, are the key forces in building the modern civilization of the Chinese nation. The deep empowerment of science and technology to culture is always the key link to realize the creative transformation and innovative development of Chinese excellent traditional culture, and it is also an important path to deal with the genes of the times, historical significance and modern value of traditional culture. In the future, the research group will continue to pay attention to new hotspots, new trends and new trends in the field of culture and science and technology, and provide intellectual support for the deep integration of culture and science and technology and the high-quality development of cultural industries through technological innovation.
Reporting/feedback
