Key Developments in China’s AI Industry

According to the “2024 World Artificial Intelligence Legal Blue Book” released at the conference, China’s core AI industry reached a scale of 578.4 billion RMB in 2023, a year-on-year growth of 13.9% according to TMT Post.

Despite a decrease in the number of investment and financing events by 18.2% compared to 2022, the total financing amount surged by 51%, totaling 263.1 billion RMB.

China’s AI industry has seen a significant number of investments, primarily from industrial investors such as Qiming Venture Partners, Sequoia Capital, Shunwei Capital, Alibaba, Tencent, and Baidu.

As of December 14, 2023, the Chinese AI industry recorded 10,110 investment events with a total financing amount of 377.62 billion RMB. The primary investment areas include enterprise services, advanced manufacturing, and automotive transportation.

Global AI Landscape

Globally, there are approximately 30,000 AI enterprises, with the US accounting for 34% and China 15%, totaling over 4,500 companies in China. The period from 2023 to the first quarter of 2024 saw 234 AI unicorns globally, with the US hosting 120 and China 71.

The report indicates that there are currently 1,328 AI large models worldwide, with the US leading at 44% and China at 36%. This positions China as a significant player in the AI domain, with a robust development trajectory.

Expert Insights and Predictions

At the WAIC 2024, several experts provided their insights into the future of AI.

Tao Mei, a foreign academician of the Canadian Academy of Engineering and founder and CEO of Zhixiang Future, emphasized the rapid growth of the AI industry.

Mei predicted that generative AI would contribute to a 14% increase in global GDP, with 74% of the world’s economic output being impacted by AI technology. He also forecasted that 26% of China’s GDP growth would be related to generative AI.

Mei likened generative AI to fundamental infrastructure like water and electricity, suggesting that it will revolutionize human-computer interaction and various sectors, including AIGC and the metaverse.

The Emergence of AI Super Apps

Li Xuexia, co-founder and CEO of Wu Wenxin, described AI large models as “super models” and AI-native applications as “super apps.”

The discussion at the conference highlighted the growing anticipation for AI super apps akin to WeChat, TikTok, and Toutiao. However, industry leaders urged patience, noting that the development of such applications involves multiple factors beyond just technological advancements.

Zhang Peng, CEO of Zhipu AI, emphasized the importance of taking action rather than merely anticipating the arrival of super apps. He pointed out that the maturity of technology, market readiness, and demand discovery are all crucial elements in the emergence of these applications.

Jun Jie, founder and CEO of MiniMax, projected that it would take at least three years for AI super apps to become mainstream, underscoring the need for a step-by-step approach to achieving this goal.

The Future of Generative AI

Zhou Zhifeng, managing partner at Qiming Venture Partners, shared his expectations for the future of generative AI at the forum. He predicted that within three years, AI video generation technology would see widespread adoption, transforming industries such as film, animation, and short videos.

Other key forecasts included:

• The gradual fusion of GPT and diffusion models, unlocking new capabilities.
• Significant improvements in high-quality data acquisition and organization, with synthetic data playing a larger role in pre-training.
• Advances in multi-agent technology enhancing the efficiency and effectiveness of generative AI.
• The emergence of unified continuous representation of images and text, leading to more powerful multi-modal models.
• A five-fold increase in the compression rate of image and video latent space representation, accelerating generation speed.
• The rise of super multi-modal large models incorporating diverse modalities like text, images, voice, music, 3D, and sensor data.
• A reduction in the cost of commanding machines to complete complex tasks, driven by AI’s ability to bridge human and machine languages.
• Significant growth in edge-side inference, propelled by advances in inference optimization algorithms, edge-side inference chips, and large models.

Zhou concluded by emphasizing the transformative potential of AI in various industries, advocating for patience as the world anticipates the advent of AI super apps.

Conclusion

China’s AI industry is on a trajectory of significant growth and innovation. With substantial funding and strategic investments, the sector is poised to make transformative impacts across various fields.

As experts predict the emergence of AI super apps and advancements in generative AI, the global AI landscape will continue to evolve, driven by technological breakthroughs and strategic foresight. The future of AI promises to be a dynamic and integral part of economic and technological development.

Operational Highlights

AI and Cloud Services

CEO Robin Li noted the stability of Baidu Core’s online marketing revenue and growth in AI Cloud revenue driven by the optimization of their AI technology stack. The ERNIE family of models has been expanded with several lightweight versions, enhancing affordability and efficiency for various applications.

Intelligent Driving

Baidu’s Apollo Go, the autonomous ride-hailing service, provided approximately 826,000 rides in Q1 2024, a 25% increase year-over-year. As of April 2024, Apollo Go has provided over 6 million cumulative rides.

Mobile Ecosystem

In March 2024, Baidu App’s monthly active users (MAUs) reached 676 million, a 3% increase year-over-year. Managed Page accounted for 50% of Baidu Core’s online marketing revenue, demonstrating significant user engagement.

AI Developer Community

The PaddlePaddle developer community grew to 13 million by mid-April 2024. Baidu launched and improved tools on its MaaS (Model as a Service) platform, facilitating the development of AI-native applications and customized models for enterprise customers.

Financial Highlights

• Revenue: Total revenues for Q1 2024 were RMB31.5 billion (US$4.37 billion), a 1% increase year-over-year.
  • Revenue from Baidu Core was RMB23.8 billion (US$3.30 billion), a 4% increase year-over-year, driven by a 3% rise in online marketing revenue
  • 6% increase in non-online marketing revenue, primarily from the AI Cloud business.
• Operating Income: Operating income was RMB5.5 billion (US$760 million), a 10% increase year-over-year.
  • Baidu Core operating income was RMB4.5 billion (US$629 million), with a 19% operating margin.
  • Non-GAAP operating income was RMB6.7 billion (US$924 million), with Baidu Core contributing RMB5.6 billion (US$774 million).
• Net Income: Net income attributable to Baidu was RMB5.4 billion (US$755 million), a 6% decrease year-over-year.
  • Non-GAAP net income attributable to Baidu was RMB7.0 billion (US$971 million), reflecting a 22% increase year-over-year.
  • Non-GAAP diluted earnings per ADS were RMB19.91 (US$2.76), up 24% year-over-year.
• EBITDA: Adjusted EBITDA was RMB8.2 billion (US$1.14 billion), maintaining a 26% margin. Adjusted EBITDA for Baidu Core was RMB7.1 billion (US$986 million), with a 30% margin.

Since the beginning of Q1 2024, Baidu has returned US$229 million to shareholders, bringing the total repurchase to US$898 million under the 2023 share repurchase program.

Unified Branding and New Releases

One of the most notable changes is the rebranding of ByteDance’s AI models. The previously known “Lark” models are now unified under the “Doubao Model Family” name.

This family includes nine models, streamlined into two primary categories: general-purpose models and seven specialized models tailored for various functions, including role-playing, speech recognition, speech synthesis, voice cloning, and text-to-image generation.

The Doubao general-purpose models are available in two versions: Pro and Lite, catering to different user needs. This approach contrasts with competitors’ broad, all-encompassing models, highlighting Doubao’s more focused strategy.

Impressive Performance Metrics

In terms of performance, Doubao models are impressive, according to data shared at the conference.

Currently, these models process an average of 120 billion tokens of text and generate 30 million images daily.

The Doubao App, a major AI application built on these models, leads the domestic AI-generated content (AIGC) market, topping download charts on both the Apple App Store and major Android platforms in China.

According to Zhu Jun, ByteDance’s Vice President of Products and Strategy, the app has seen the creation of over 8 million intelligent entities and boasts 26 million monthly active users.

Competitive Pricing Strategy

Volcano Engine also announced a groundbreaking pricing strategy, setting the cost at 0.0008 yuan per thousand tokens. This is significantly lower than competitors, such as OpenAI’s GPT-4o, which charges approximately 0.035 yuan per thousand tokens for input and 0.1 yuan per thousand tokens for output.

In comparison, domestic large model providers average around 0.12 yuan per thousand tokens.

This substantial price reduction is attributed to several factors, including optimization in model structure and the adoption of distributed inference methods, which enhance computing efficiency.

Volcano Engine’s head, Tan Dai, emphasized that this pricing strategy is not intended as a price war but follows a sustainable business logic aimed at fostering long-term commercial cooperation.

AI Market and Future Developments

Despite the advancements, the AI application market is still in its early stages.

According to QuestMobile, the AIGC industry had 73.8 million users as of March 2024, a significant increase from the previous year but still only 6% of the total mobile internet users.

The competitive landscape is gradually intensifying, with companies like Alibaba Cloud and Tencent Cloud also announcing price cuts, although primarily for CPU-based general computing power rather than AI-specific GPU computing power.

ByteDance’s aggressive pricing for AI capabilities is a strategic move to lower the barrier for AI adoption and expand market reach.

This strategy is expected to continue throughout the year as foundational model prices further decline, driven by technological advancements and increased market competition.

Human-like AI and Accelerated B2B Adoption

The trend towards more human-like AI is evident in this year’s major tech releases. OpenAI’s GPT-4o and Google’s Project Astra have set new standards for interactive AI, with real-time conversational capabilities and emotional intelligence.

Similarly, ByteDance’s Doubao App aims to provide a more human-like interaction experience, with enhanced ASR (Automatic Speech Recognition) and TTS (Text-to-Speech) technologies, making AI interactions feel more natural and intuitive.

In the B2B sector, ByteDance’s AI ecosystem is rapidly taking shape. Volcano Engine is building a robust developer ecosystem and AI services platform, supporting a wide range of applications from mobile phones to smart vehicles.

Strategic partnerships with major automakers like Geely, Great Wall Motors, and Seres, along with integration into smartphones from Oppo, Honor, and Xiaomi, highlight ByteDance’s comprehensive approach to AI adoption.

Volcano Engine has also launched the enterprise version of its low-code AI development platform, “Coze” (Kouzi), offering high flexibility and various integration options for developers.

This platform is already being utilized by leading enterprises such as China Merchants Bank, Haidilao Hot Pot, and Super Monkey fitness centers to build intelligent assistants and personalized services.

The user base has exceeded 200 million, with daily API calls also surpassing 200 million. The service now caters to 85,000 customers and has more than 190,000 AI-native applications developed using the Qianfan platform.

Robin Li also noted that one doesn’t need to know coding to create certain applications; it’s possible to develop smartphone apps without programming. AI is fostering a wave of creative energy. “The future of application development will be as simple as making a short video; everyone will be a developer,” Li stated.

After officially announcing the open access to “Ernie Bot” in August last year, Baidu has on one hand used large models to rebuild client-side products for consumers and integrated AI features across multiple Baidu products. On the other hand, in the business sector, it has launched the enterprise-level one-stop large model platform “Qianfan Large Model Platform” through its cloud business.

Baidu’s GenAI App Offers Digital Twin Creation

In early 2024, Baidu has announced an innovative feature within its Wenxin Yiyan app (its GenAI product; a.k.a. Ernie Bot) that allows users to create digital twins with unprecedented ease. Both iOS and Android users can access this feature for free upon updating to the latest version of the app.

The new function simplifies the creation of a digital twin to just a photo and three voice recordings. Users can customize their digital counterpart with a unique name, voice, and even MBTI personality type. This flexibility extends to the choice of making their digital twin public or keeping it private, catering to various needs such as companionship, knowledge exploration, and entertainment.

Traditionally, creating a digital twin involved complex processes like 3D scanning to create a model, texturing through rendering technologies, and integrating motion capture and voice synthesis. Baidu’s latest update significantly reduces the complexity and time required to create a digital persona.

Baidu also teased upcoming themes for digital twins, including various New Year backgrounds and exclusive digital personas for a “mysterious national idol,” AI God of Wealth, and AI family portraits among other activities.

Recently at the tenth Wave Summit, a biannual deep learning developer conference hosted by Baidu, the Chinese AI giant unveiled remarkable achievements in artificial intelligence.

Ernie Bot has reached a new milestone, surpassing 100 million users as of Dec 2023.

ERNIE Bot: A Versatile AI Tool

Since its public release on August 31, 2023, ERNIE Bot has rapidly become an indispensable tool for users across diverse fields. It has generated 3.7 billion words of text in workplace environments, aiding over 2 million users, according to information CIW received from Baidu.

ERNIE Bot’s versatility is further highlighted by its ability to write 300 million lines of code and process 400 million words of contracts, showcasing its utility in technical and legal domains. Additionally, it has created 5 million travel itineraries and drafted 10.83 million messages of care and support, reflecting its wide-ranging applications in personal and emotional contexts.

The popularity of ERNIE Bot is also reflected in the user feedback, with 20 million likes and appreciations, signifying high satisfaction and reliance on this AI tool.

Technical Innovation and Industry Empowerment

The Wave Summit also shed light on Baidu’s commitment to technical innovation and industry empowerment through its PaddlePaddle platform. Dr. Haifeng Wang, Chief Technology Officer of Baidu, revealed that PaddlePaddle has been instrumental in supporting over 10.7 million developers and 235,000 enterprises.

A staggering 860,000 models were created using the platform as of December 2023, indicating the platform’s significant impact on AI development and application.

The Future of ERNIE Bot and Baidu’s AI Initiatives

The continuous improvement of the ERNIE Foundation Model, which saw a 32% overall performance enhancement in just two months, clearly indicates Baidu’s dedication to advancing AI technology.

ERNIE Bot’s growing popularity and increasing user inquiries highlight the potential and demand for advanced AI solutions in various industries.

Baidu’s achievements with ERNIE Bot and the PaddlePaddle platform underscore the company’s position as a leader in China’s AI innovation. As Baidu continues to break new ground in AI technology, the global community eagerly anticipates further advancements and applications that will transform industries and everyday life.

originality.ai’s founder and CEO, Jonathan Gillham, recently wrote an in-depth analysis article discussing Baidu’s ERNIE Bot and its latest model- ERNIE 4.0, which you can find here.

AI-Enhanced Automotive Assistants

NIO’s incorporation of AI through Microsoft Azure OpenAI service into its NOMI assistant is not just an upgrade—it’s a transformation.

NOMI now employs a car-specific version of GPT (Generative Pre-trained Transformer) to handle complex queries and provide intuitive, vehicle-specific information. This enables drivers to receive more accurate and relevant responses without diverting their attention from driving, enhancing safety and user experience.

The AI capabilities include personalized greetings, weather updates, and a new range of interactive features designed to optimize journey experiences by providing timely operational guidance for vehicle features.

For example, when asked, NOMI can now explain how to adjust mirrors or check tire pressure directly via the car’s display—tasks traditionally requiring manual lookup in user manuals.

Interactive Features and User Engagement

Beyond the basics, NOMI GPT introduces enhanced interactivity through a new FAQ feature that educates users about NOMI’s capabilities and preferences, personalizing the user experience further.

This feature allows users to engage in more natural conversations with their vehicle, asking about NOMI’s favorite music, games, or even its opinions on other AI systems.

Phased Rollout and Language Support

To ensure the best and safest user experience, NIO has begun a phased rollout of the NOMI GPT feature, starting from April 5, 2024.

The feature is initially available in English, German, and Norwegian, catering to NIO’s diverse European market. This strategic deployment underscores NIO’s commitment to accessibility and customer satisfaction through its leading over-the-air (OTA) update capabilities.

Implications for the Future

The integration of advanced AI into NIO’s vehicles is more than a mere enhancement; it’s a forward-looking move that signals the future direction of the automotive industry.

As articulated by NIO’s European Product Experience Director, Benjamin Steinmetz, the new AI features not only make the user’s journey more valuable but are also set to expand quickly, bringing exciting new functionalities with each update.

This development is not just about enhancing individual vehicles but about pushing the entire industry toward smarter, more interactive, and more intuitive automotive solutions.

The partnership between NIO and Microsoft, through the use of Azure and OpenAI technologies, exemplifies how collaborations between tech giants and automotive innovators can drive significant advancements.

Conclusion

The evolution of AI in vehicles through initiatives like NIO’s enhanced NOMI assistant demonstrates the potential for AI to transform everyday experiences and interactions with technology.

As these technologies mature and become more integrated into various aspects of driving and vehicle management, they pave the way for more connected, intelligent, and user-centric automotive environments.

This trend is set to continue, reshaping how we think about mobility and technology’s role in it, with China at the forefront of this technological revolution in the automotive sector.

Focusing on the Chinese market, investments in AI are projected to hit $38.1 billion by 2027, accounting for nearly 9% of the global investment pool.

In recent years, rapid technological advancements and market maturity have spurred Chinese companies to accelerate their intelligent upgrades. The emergence of large AI models at the end of 2022 provided a crucial overtaking opportunity for new entrants to the market.

Previously, the widespread adoption of large models was limited, creating bottlenecks as companies sought to transition AI technologies from pilot trials to full-scale applications, hampered by a fragmented talent and resource landscape.

The introduction of large AI models has significantly lowered the barriers to technology adoption and enhanced AI’s generalization capabilities, fostering the development of innovative applications and solutions. The model-as-a-service business model could potentially reshape the current market dynamics, leading China into a new era of digital business development.

IDC’s recent release, “IDC FutureScape: Worldwide AI and Automation 2024 Predictions – China Implications,” offers vital insights for IT and business decision-makers responsible for leveraging solutions and directing AI and automation expenditures.

The report, outlining ten predictions for the AI and automation markets from 2024 to 2029, assesses each forecast based on its impact and time frame. These predictions are designed to guide strategic planning, adjustments, and resource allocation in the AI and automation spheres for IT and business leaders.

Key Predictions for China’s AI and Automation Market

1. GenAI Security and Governance: By 2025, 40% of cloud and software platform providers will integrate GenAI security and governance with their primary services, tripling the reduction in GenAI-related risks.
2. Regulatory Divergence: AI regulatory differences across regions will pose significant challenges for multinational companies by 2026, increasing the time and effort required to manage sensitive issues by 20%.
3. Digital Assistants Over Software Interfaces: By 2027, GenAI digital assistants will replace 20% of enterprise software interfaces, playing an increased role in business processes.
4. Outcome-Focused Automation: Following GenAI technology attention, 50% of enterprises will focus on the outcomes of GenAI rather than the implementation of specific technologies by 2024.
5. AI in IT Expenditure: By 2026, China’s top 500 companies will allocate over 30% of their core IT spending to AI, accelerating product innovation and process improvement at a double-digit rate.
6. Stabilizing Economic Impact of AI: By 2028, the overall economic impact of AI will stabilize as companies address fundamental issues and refocus resources on innovation and new business opportunities, promoting steady growth.
7. Practical Application of Automated Knowledge Retrieval: Two-thirds of enterprises will use a combination of generative AI and RAG for domain-specific automated knowledge retrieval by 2027, improving decision efficiency by 55%.
8. Disruptive Business Models: By 2025, 50% of China’s top 500 companies will leverage innovative business models to double the profit potential of generative AI.
9. Impact of AGI: By 2027, 25% of China’s top 500 companies will experiment with AGI systems, heralding transformative societal impacts along with significant opportunities and challenges.
10. Reversal in Chip Revenue: By 2028, spending on accelerator chips (GPUs, FPGAs, ASICs, or ASSPs) versus traditional CPUs will invert, reaching a 55/45 ratio.

China’s AI Market Booms with Focus on Professional Services, Government, and Finance