In-depth Research on the AIGC Industry Market and Comprehensive Analysis of China's AIGC Industry Status

[baoshi.rao]

Analysis of the Concept and Development Stages of AIGC

Generative AI (AIGC) is a pivotal milestone in the evolution from AI 1.0 to AI 2.0. The integration of technologies like GAN, CLIP, Transformer, Diffusion, pre-trained models, multimodal techniques, and generative algorithms has catalyzed the rapid advancement of AIGC. Continuous algorithmic innovation, the qualitative leap in AIGC capabilities driven by pre-trained models, and the expansion of AIGC content through multimodal approaches have endowed AIGC with more versatile and robust foundational capabilities.

From the progression of computational intelligence to perceptual intelligence and then to cognitive intelligence, AIGC has opened the door to cognitive intelligence for human society. By training on large-scale datasets, AI can acquire knowledge across multiple domains, enabling it to perform tasks in real-world scenarios with appropriate model adjustments. Generative AI—AIGC—refers to the technology that leverages methods like generative adversarial networks (GANs) and large pre-trained models to generate relevant content through data learning and recognition, with suitable generalization capabilities. The core idea of AIGC is to use AI algorithms to produce creative and high-quality content. By training models on vast datasets, AIGC can generate content based on input conditions or guidance. For example, given keywords, descriptions, or samples, AIGC can produce matching articles, images, audio, and more.

AIGC holds landmark significance for human society and AI. In the short term, it transforms foundational productivity tools; in the medium term, it reshapes societal production relations; and in the long term, it drives qualitative breakthroughs in overall social productivity. In this transformation of productivity tools, production relations, and productivity, the value of data as a production factor is greatly amplified. AIGC elevates data to the status of a core resource of the era, accelerating the digital transformation of society.

With the continuous iteration of deep learning models, AIGC has achieved groundbreaking progress. Particularly in 2022, algorithms experienced explosive growth, and breakthroughs in underlying technologies made AIGC commercialization feasible. Key developments include:

• 2014: Introduction of Generative Adversarial Networks (GANs).
• 2021: OpenAI's release of the CLIP (Contrastive Language-Image Pre-Training) multimodal pre-trained model.
• 2022: Diffusion models gradually replacing GANs.

Before 2021, AIGC primarily generated text, but newer models can handle formats like text, speech, code, images, videos, and robotic actions. AIGC is regarded as a new content creation paradigm following PCG and UGC, leveraging its advantages in creativity, expressiveness, iteration, dissemination, and personalization. Post-2022, AIGC has grown rapidly, fueled by improvements in deep learning models, open-source initiatives, and the exploration of commercialization for large models, serving as the "accelerator" for AIGC development.

Analysis of the Current State of the AIGC Industry

AIGC is an AI technology built on multimodality, enabling a single model to understand language, images, videos, and audio simultaneously and perform tasks beyond the reach of unimodal models, such as adding captions to videos or generating images based on semantic context. Currently, domestic AIGC applications mostly appear as single-model solutions, categorized into text generation, image generation, video generation, and audio generation, with text generation serving as the foundation for other content types.

The AI industry chain consists of three layers:

1. Infrastructure Layer: Focuses on foundational support platforms, including sensors, AI chips, data services, and computing platforms.
2. Technology Layer: Centers on core technology R&D, covering algorithm models, foundational frameworks, and general technologies.
3. Application Layer: Emphasizes industrial applications, including industry-specific solutions, hardware, and software products.

China's AIGC industry chain comprises five components: foundational large models, industry/scenario-specific medium models, business/domain-specific small models, AI infrastructure, and AIGC support services, forming a robust ecosystem.

Globally, AIGC commercialization began with foundational large models, exemplified by applications like ChatGPT and Midjourney, which are built on such models. In contrast, China's commercialization started with business/domain-specific small models due to its highly diverse business scenarios and fragmented supply-side services. Foundational large models in China are still in rapid iteration but are increasingly focusing on specific business scenarios. Industry/scenario-specific medium models lag behind but represent a future area of cross-domain collaboration between foundational large models and domain-specific small models in China's unique market.

In February 2023, Meta CEO Mark Zuckerberg announced the formation of a top-tier product team dedicated to AIGC, while YouTube revealed plans to develop AIGC content creation tools. Overseas giants like OpenAI, Microsoft, Google, and Buzzfeed have already deployed related services. Domestically, companies like BAT, ByteDance, Wanxing Technology, and BlueFocus have swiftly expanded into the AIGC sector following the 2022 AI art boom.

Currently, the supply and demand for AIGC (Artificial Intelligence Generated Content) technology and applications in China are primarily concentrated in areas such as marketing, office work, customer service, human resources, and basic operations. The empowering effects and value of this technology have been preliminarily validated. According to the TE Think Tank's 'Enterprise AIGC Commercial Application Research Report,' 33% of companies urgently expect AIGC enhancement and support in marketing scenarios, 31.9% in online customer service, 27.1% in digital office scenarios, and 23.3% in information and security scenarios.

In 2023, the rise of AIGC (Generative AI) large models has also sparked an 'AI+' wave in the home furnishing industry. At Dongyi Risheng's product launch, it was revealed that their three AI-assisted home decoration applications—'Creative Master,' 'Real Home AIGC,' and 'Beginner Designer'—have opened up new perspectives on AIGC in the home decoration industry and introduced a new era of intelligent home decoration experiences. Additionally, Shangpin Home Collection also released AIGC technology based on multimodal large models in May.

On September 7, UNESCO issued its first guidelines on the use of generative AI (AIGC) in education, urging government agencies to regulate the technology, including protecting data privacy and setting age restrictions for users. UNESCO Director-General Audrey Azoulay stated that generative AI could be a tremendous opportunity for human development but might also cause harm and bias. She added that without public participation and necessary government safeguards and regulations, it cannot be integrated into education.

Future Trends in the AIGC Industry

Data shows that China's core AI industry scale reached 150 billion yuan in 2020 and is expected to grow to 400 billion yuan by 2025, potentially making it the world's largest AI market. AIGC will greatly unleash human imagination and usher in a 'new art wave' for this era. AIGC saw explosive growth at the end of 2022, with the intelligence level of AI models continuously rising due to the interplay of data, algorithms, and computing power, gradually awakening AI's 'creative intelligence.' As AI technology continues to advance and break through, generative AI is accelerating its penetration into fields such as text, images, and audio-video content.

With the continuous development of AIGC technology, its role in the AI field will become increasingly important. In the future, AIGC will be applied in more areas, such as smart homes and intelligent transportation. At the same time, AIGC will also face more challenges, such as data security and privacy protection, requiring continuous technological innovation and refinement.