Robin Li on AI: Domestic AIGC Applications Enter a New Phase of Orderly Development

[baoshi.rao]

On the morning of the 9th, the 2023 World Internet Conference Wuzhen Summit Internet Entrepreneurs Forum was held. Robin Li, founder, chairman, and CEO of Baidu, mentioned in his speech that the ultimate goal of technological innovation is application. The hallmark of humanity entering the AI era is the emergence of a large number of AI-native applications, not just large models. Large models can support the development of countless AI-native applications, but only with millions of AI applications can large models be considered successful.

According to a report on September 20, the popular AI application 'Miao Ya Camera' announced the launch of a free version, allowing users to generate personalized digital avatars by uploading 8 photos.

The report stated that with enhanced computing power and optimized algorithms, the efficiency of Miao Ya Camera in generating digital avatars has significantly improved, now taking an average of about 20 minutes. Initially, the waiting time ranged from 1 to 3 days.

Zheshang Securities noted that for AI applications like 'Miao Ya,' soft technologies possess the characteristic of 'zero marginal cost expansion,' meaning the penetration rate of such applications is expected to grow exponentially. They also pointed out that as the AI industry wave deepens, the profit and valuation potential of soft technologies, represented by computer and media sectors, are expected to expand further.

Changjiang Securities believes that for Miao Ya, computing power support and its costs remain the core constraints in meeting user demand promptly. Computing power is still the most certain and clearly demanded core track under the current industry trend. According to Huajin Securities' estimates, training computing power in 2022 was 95 Eflops, projected to reach 749 Eflops by 2025, with a compound growth rate of 43%.

Regarding companies, Guojin Securities and GF Securities research reports indicate that Wanxing Technology has launched the overseas version of Miao Ya, 'Pixpic,' which requires 10 photos to generate a digital avatar and is optimized for the European and American user base. Visual China has also ventured into AI-generated images with its AI-inspired drawing layout.

AIGC Technology Market Size Research

AIGC will serve as a productivity tool to drive the development of the metaverse and further promote the broader application of AI technology.

From the perspective of metaverse growth, whether at the government level or among investment institutions, the metaverse has accumulated significant momentum. Since 2022, major companies like Huawei, Samsung, and Sony have continued to invest in AR/VR terminals and various devices, anticipating that new product launches will enhance user experience and boost hardware penetration.

The future metaverse market size is expected to reach $600 billion, with unified international technical standards providing better support for its development. Additionally, as metaverse commercialization accelerates, AI will play a crucial role in creating digital humans, virtual scenes, and virtual assets.

Regions such as Beijing, Jiangsu, Shanghai, Zhejiang, Shandong, and Fujian have密集 released development plans, measures, and action plans for the metaverse and related industries, competing for early advantages in the metaverse industry.

The '14th Five-Year Plan for Digital Economy Development' outlines China's digital economy goals: the added value of core digital economy industries is expected to increase from 7.8% of GDP in 2020 to 10% by 2025.

Currently, companies like Baidu, Tencent YouTu, Alibaba, Kuaishou, ByteDance, NetEase, SenseTime, Meitu, and iFlytek are increasing their investments in AIGC. Zheshang Securities stated that the AI industry is accelerating its transition from cutting-edge technology exploration to commercial implementation, with rising industry momentum. Leading companies' active participation in application competition will accelerate the integration of AIGC applications across the industry.

Moreover, AIGC-powered music is one of the core focuses for Tencent Music's future efforts. Liang Zhu mentioned that the company is also using AIGC innovations to provide highly personalized and differentiated music entertainment experiences, meeting users' demands to keep up with the AIGC wave, thereby enhancing the platform's appeal.

In online music, the company has begun testing 'AI Listen Together,' where the AI companion 'Xiao Qin' interacts with users, shares topics, and recommends songs and playlists in real-time.

Domestically, the policy environment continues to optimize, with strong demand for virtual humans and human-computer interaction scenarios, which is expected to accelerate the application of AI technology in fields such as finance and healthcare. The AIGC concept has also become a hot topic on investor relations platforms, with several listed companies, including Silu Vision and Jihong Shares, recently disclosing related layouts.

The 2023-2028 China AIGC Industry Market In-depth Research and Development Strategy Report published by the China Research Institute shows that AIGC has already begun and is expected to be widely used in the creation and generation of text, images, music, videos, and game materials. AIGC may become a catalyst for improving the supply and commercialization efficiency of digital content in the metaverse.

With its characteristics of authenticity, diversity, controllability, and combinability, AIGC is expected to help companies improve the efficiency of content production and provide more diverse, dynamic, and interactive content. It is likely to achieve significant innovative developments first in industries with high digitalization and rich content demands, such as media, e-commerce, film, and entertainment.

ChatGPT is a phenomenon-level application in the AIGC industry. Released by OpenAI on November 30, 2022, ChatGPT is a general-purpose dialogue system that can perform tasks such as programming, Q&A, and text generation through simulated conversations. Its sustained popularity has opened up broad space for subsequent commercialization and application, as well as providing vast development opportunities for cognitive intelligence technologies centered on natural language processing. NVIDIA CEO Jensen Huang stated that 'ChatGPT is like the iPhone moment for AI.'

The AIGC industry has high entry barriers, with existing companies controlling key resources to a significant extent. However, the industry is growing rapidly and is highly attractive, making the overall threat from new entrants moderate. The main threats come from internal product and service substitutions and competition, as well as updates and iterations between different technologies. Overall, the threat of substitutes in China's AIGC industry is relatively small.

Due to rapid technological iterations, large market demand, and numerous sub-sectors, China's AIGC industry has many participants, with leading internet companies holding key resources. Overall, competition among existing companies is intense.

China's AIGC applications have entered a new phase of orderly development. The Cyberspace Administration of China and six other departments issued the Interim Measures for the Management of Generative Artificial Intelligence Services (hereinafter referred to as the 'Measures'), which impose specialized regulations on generative AI and came into effect on August 15, 2023.

Generative AI (AIGC) technology is developing rapidly and is entering our lives and work through various software and terminal devices. However, while AIGC brings convenience, it also poses legal risks.

1. Overseas AIGC service providers (whether at the model or application layer), whether directly providing services to China or indirectly through APIs or other forms of 'packaging,' will be subject to the relevant provisions of the Measures.

In scenarios such as chat dialogues, where there are both real users and AI chatbots, it is necessary for machine review to quickly identify harmful information to ensure the real-time nature of human-machine conversations.

AI machine review technology can configure different levels of review strategies for user-generated content (UGC) and AIGC-generated content based on business scenarios, balancing user experience with content safety and compliance.

2. AIGC service providers are required to fulfill an exceptionally high obligation: 'ensuring the authenticity, accuracy, objectivity, and diversity of data.' Considering the practical realities of AIGC technology development, the management measures have been reasonably adjusted to 'take effective measures to improve the quality of training data, enhancing its authenticity, accuracy, objectivity, and diversity.'

3. 'Foreign investment in generative AI services must comply with relevant foreign investment laws and regulations.' Given the current lack of clarity in AIGC-related foreign investment rules, it is advisable to stay updated on legislative developments. It is anticipated that different foreign investment restrictions will apply based on future classification and tiered regulatory rules.

AIGC Enterprise Competitive Landscape

Currently, many brand companies are exploring virtual trial applications, such as Uniqlo's virtual fitting rooms, Adidas' virtual shoe trials, Chow Tai Fook's virtual jewelry try-ons, Gucci's virtual watch and glasses trials, IKEA's virtual furniture matching, and Porsche's virtual test drives. Although traditional manual modeling methods are still used, advancements in AIGC technology are expected to lower the barriers and costs of 3D modeling, enabling large-scale commercial use of virtual try-on applications.

Market share analysis includes companies like Huawei, Baidu, Tencent, and the emerging unicorn Xiaoice. Among China's leading AIGC enterprises, internet giants have the broadest business布局, primarily in first-tier and new first-tier cities, with extensive strategic布局 across various sectors. However, startups also have their advantages, achieving notable success in niche markets.

AIGC Market Share Reached Only About 20% in 2023

Currently, China's AIGC industry has low market concentration, with intense product competition. Based on the concentration of AI patent applications, as of May 2023, the top four companies accounted for only about 20% of the market share, while the top eight accounted for approximately 25%, indicating low market concentration.

The low market concentration is primarily due to the rapid迭代 of AI technology, where even leading companies cannot maintain a comprehensive technical advantage. Additionally, the AIGC industry has many细分赛道 and is still in its early stages, offering vast market potential.

AIGC Market Size Forecast (2030)

The penetration rate of AIGC in content generation is expected to rise rapidly, with applications expanding quickly. By 2030, the AIGC market size is projected to exceed one trillion yuan.

With active布局 by industry giants, accelerated industrial落地, and共振 from AI technological innovations (e.g., generative algorithms, pre-trained models, multimodal technologies), industrial ecosystems (a three-tier ecosystem雏形已现), and policy support (e.g., Beijing's support for头部企业 developing large models对标ChatGPT), AIGC is poised for rapid growth. According to Tencent Research Institute's AIGC trends report, the market size is expected to reach $110 billion by 2030, showcasing vast prospects.

The AIGC industry research report aims to预测 future business market prospects based on national economic and industrial development strategies, helping clients navigate policy uncertainties and identify investment opportunities. Through extensive analysis and预测, the report studies future development and investment strategies, enabling companies to gain insights in a competitive market and adjust their经营策略 according to AIGC industry demands.