As technology giants compete for big AI models, artificial intelligence technology has become the most watched new technology this year. However, the road to innovation in large AI models still lacks a sufficiently powerful commercial answer.
As the world's first major Internet company to release large models, whether Baidu can accelerate the application of large models and how much commercial value Wenxin large models can create have always been the focus of market attention.
On the evening of November 21, Baidu released its third quarter results for 2023. According to the financial report, Baidu's third-quarter revenue reached 34.447 billion yuan, and net profit (non-GAAP) attributable to Baidu reached 7.3 billion yuan, a year-on-year increase of 23%. Both revenue and profit exceeded market expectations.
Taking a closer look at the performance growth in the financial report, the big model undoubtedly stands in the "C position".
On the one hand, with AI-native thinking, Baidu accelerates the integration of large models with existing businesses to empower business innovation. After the release of the financial report, Baidu founder, chairman and CEO Robin Li said, “In the context of emerging opportunities in generative AI and basic models, Baidu continues to strengthen its technology and improve its products, especially the launch of Wenxin Big Model 4.0. This is Baidu's most powerful basic model. We have also used Wenxin's large model and Wenxin Yiyan's capabilities to reconstruct consumer-level and enterprise-level products and our own operations, bringing AI-native experiences while continuously improving efficiency. "
On the other hand, Wenxinyiyan continues to evolve efficiently and connects 70 million users and millions of ecological partners, laying the foundation for sustainable growth. In this regard, Robin Li pointed out: "We have fully opened the Wenxin Big Model API to cloud customers to help them develop AI native applications and solutions. Baidu adheres to the business and product strategy with artificial intelligence as the core, providing Wenxin Big Model and Wenxin Yiyan Ecology has laid the foundation for revenue maintenance and profit expansion in the coming years."
Based on Baidu's recent development trends, whether from the internal perspective of its own business transformation or the ecological perspective of external empowerment, Baidu has leapt across the "innovation canyon" between AI large model technology development and commercial transformation. The bottom layer of its growth The logic is changing.
Reconstruct traditional businesses and expand mobile ecological boundaries with large models
After the financial report was released, Baidu Chief Financial Officer Luo Rong pointed out that Baidu Core maintained stable profit margins during the quarter and that Baidu's continued investment in artificial intelligence has laid the foundation for technology and product innovation. Going forward, Baidu will continue to prioritize investment in artificial intelligence, especially generative AI and basic models, and Baidu will unswervingly focus on efficiency and strategic resource allocation.
From his words, the American Stock Research Institute understands that Building a new growth engine around generative AI and large language models is Baidu’s main evolutionary direction in the future.
This means that native applications based on large AI models will play an increasingly important role in Baidu's business. Internally, Baidu uses the Wenxin large model to reconstruct its own business; externally, Baidu uses large models to empower hundreds of industries and explore in-depth applications in vertical fields.
First of all, from an internal perspective, AI native reconstruction is the key word.
Since the release of Wenxinyiyan’s basic technical architecture in March, Baidu has continued to use AI-native thinking to upgrade and transform traditional businesses. At the Xili Lake Forum in Shenzhen on November 15, Robin Li, founder, chairman and CEO of Baidu, said: "Baidu has resolutely restructured its product lines to be AI-native. For every 100 lines of code on Baidu, 20 lines were generated by AI."
This restructuring is reflected at the product level, manifesting as breakthrough progress in user experience and business growth. Among them, the most obvious one is Baidu's "basic market" - mobile ecology.
On the one hand, AI native applications improve the user experience and business efficiency of various products in the mobile ecosystem. For example, Baidu's "New Search" can generate multi-modal answers in text, pictures, and dynamic charts when users search for questions. The conversational "Search" can also target complex needs in multiple Provide more accurate and personalized content in round-robin interactions.
Source: Baidu search
A glimpse of the leopard shows that behind this change in search methods is the comprehensive "reconstruction" of Baidu's business system by the big model, which has brought new productivity and imagination.
Against this background, the reconstructed AI native applications have achieved significant growth in the number of users: In Baidu App, 640,000 creators have used AI-assisted creation tools to produce 14 million pieces of content and received 30 billion views; Baidu Netdisk After "Yun Yiduo" was launched, it became the world's first "personal cloud intelligent assistant" with 20 million users. The cumulative number of users using Baidu Wenku's AI new functions exceeded 30 million. In September, the total number of members who actively activated the payment rate increased year-on-year. Over 50%.
It is foreseeable that through large AI models + data precipitation, Baidu will continue to produce new AI applications in traditional application scenarios such as search, network disk, library, and maps, which will bring about changes in overall productivity. The user stickiness and activity of Baidu AI native applications will continue to rise.
On the other hand, the value of large model reconstruction has been intuitively fed back to Baidu's business growth. Among the vertical application areas of large models, advertising is undoubtedly the area with the clearest path to commercialization. The financial report shows that Baidu’s core revenue in this quarter was 26.6 billion yuan, of which Baidu’s core online marketing revenue increased to 19.7 billion yuan.
A closer look at the reasons for the growth of Baidu's online marketing revenue is that the launch of native applications such as the AI Native marketing platform "Qing Ge", the AI digital human platform "Qing Duo" and the conversational marketing product "Brand BOT" is the key.
For example, AIGC's marketing creative platform "Qingduo" has improved the efficiency of corporate content production and advertising, and driven the continuous upgrade of customers' advertising experience.
Source: Baidu Marketing Center
For a liquor client that did not have a professional team to produce videos, after placing digital people through "Qingduo", the optimizer produced 6 digital people videos in less than an hour, significantly reducing the conversion cost and increasing the conversion rate by 12%. average consumption. The role of new AI application products in driving Baidu’s advertising business is self-evident.
AI native reconstruction is becoming the guarantee for the long-term competitiveness of Baidu’s core business. As CICC said, large models are gradually penetrating Baidu’s entire product line, accelerating the absorption of end-user feedback, promoting the optimization of model performance, and improving the ability to adapt to different scenarios; at the same time, The reconstructed product line is also expected to increase the competitive barriers of the basic business.
Of course, in addition to internally restructuring products and services and turning the traditional business flywheel, Baidu's growth also comes from its ecological moat that continues to expand externally. It can even be said that the other side of the Wenxin model - "external empowerment" better reflects the changes in Baidu's growth logic.
Empowering thousands of industries, large models shape the ecological power of business growth
“The sign that humans have entered the AI era is not the production of many large models, but the production of many native AI applications.”
Regarding the popularization of AI technology applications, Robin Li believes that AI technology should be integrated into various applications to provide intelligent and automated solutions for various industries and fields.
AI native applications will be a market worth millions. This is something obvious when compared with the mobile Internet era: there are 8 million applications developed based on Android and iOS, and the development of native applications based on large models is a long-term trend in the AI era.
What Baidu needs to do is to continue to expand its large model ecosystem around users, customers, and partners, and strengthen the company's leadership position in the era of generative AI.
Therefore, on the one hand, Baidu connects users through generative AI products and AI native applications to explore new monetization opportunities for large models; on the other hand, it empowers customers to help them seize the opportunities of generative AI.
In terms of linking users and more developers, Baidu Wenxinyiyan is releasing its own commercial monetization potential.
Since the third quarter of 2023, Baidu Wenxin Yiyan has been open to the public, and has launched Wenxin Large Model 4.0 and professional versions. As of now, Wenxinyiyan has 70 million users, covering 4,300 scenarios.
Data released at the Wuzhen Summit of the 2023 World Internet Conference (Source: Baidu AI)
Recently, the "Wenxin Yiyan Professional Edition" based on Wenxin Large Model 4.0 has been launched and is available for a fee to meet the needs of professionals. This also distinguishes Wenxin Large Model from other models on the market. Compared with the free basic version, the professional version has stronger model capabilities and image generation capabilities.
Coupled with the release of products such as Wenxinyiyan's plug-in ecosystem "Lingjing Matrix" and industrial-level deep learning platform "Flying Paddle", Wenxinyiyan continues to lead domestic peers in terms of calls. Not long ago, Robin Li said in a speech that since its opening on August 31, the number of API calls made by Wenxin Big Model has increased exponentially. The number of calls made by Wenxin Big Model alone is greater than that of more than 200 large models in China. It adds up to more.
These data mark the gradual improvement of Baidu’s open source model ecosystem and the increasingly rich model implementation scenarios.
At the same time, in terms of empowering customers, Baidu continues to make efforts in "cloud and intelligence integration".
Due to the rapidly rising demand for AI native applications, Baidu Smart Cloud provides customers with a full-stack service solution covering five major needs, including AI computing power, large model API calls, large model secondary development, AI native application development tools and AI native app Store. Currently, Baidu Intelligent Cloud Qianfan large model platform serves more than 20,000 companies, covering nearly 500 scenarios.
Judging from the specific cases in the first three quarters, enterprises' willingness to use large models has gradually increased, and more and more enterprises have joined the wave of developing AI native applications.
For example, Yulin Energy, a subsidiary of China Energy Group, cooperated with Baidu Intelligent Cloud to develop an intelligent auxiliary transportation management system for coal mines. Through the intelligent auxiliary transportation assistant built by Baidu's large model, auxiliary transportation efficiency and safety management and control capabilities continue to improve.
Image source: Baidu Smart Cloud
In addition, TAL from the education industry has completed the training and development of large models based on Baidu Intelligent Cloud; Wall-Facing Intelligence from the artificial intelligence industry has cooperated with Zhihu to train the "Zhihaitu AI" large model based on Baidu Intelligent Cloud infrastructure services. The model has a kilocalorie training monthly failure rate of less than 5%, ensuring the continuity of model training and achieving good model convergence effects.
Energy, education, technology... Baidu's "Cloud + AI" empowers multiple industries, revealing that the combination of generative AI and Baidu's existing cloud business is reshaping the distribution of cloud computing market share. In this regard, Goldman Sachs predicts that generative AI will help the steady development of Baidu Cloud business and benefit from the incremental revenue of MaaS (expected to reach 5 billion yuan in 2025). Revenue is expected from 2023 to 2025 The compound annual growth rate is 23%.
At present, for the development of large AI models, what investors value most is whether they can find a commercialization path and realize the implementation of application-side products. The application of Baidu's AI technology has been popularized and integrated into various applications, and its implementation has breadth and depth.
Therefore, CITIC Securities issued a research report stating that AI continues to transform the company's advertising business and cloud computing business, which is expected to bring about growth in user time, advertising prices, cloud infrastructure usage, and cloud native business.
The flywheel of commercialization of Wenxin large models has begun. As the results of AI's transformation of traditional businesses and external ecological construction are reflected in financial reports, Baidu is opening up its own growth space.
Conclusion: Innovators who open the door to the era of AI native applications will win in the future
Archimedes said: "Give me a fulcrum, and I can move the earth." Judging from this financial report, Baidu has found the fulcrum to leverage the commercialization of Wenxin's large model-AI native applications.
At present, AI native applications have brought incremental performance to Baidu. Net profit (non-GAAP) attributable to Baidu core reached 7 billion yuan this quarter, a year-on-year increase of 21%. This proves that expanding the AI native application ecosystem based on large models is a strategy worthy of firm and long-term advancement.
Looking to the future, with the arrival of the AI era, Robin Li's view is gradually becoming a reality: a powerful basic model will drive the explosion of AI native applications, and a prosperous AI native application ecosystem will drive economic growth. From AI technology trends to application implementation, Baidu is the industry leader. Therefore, Goldman Sachs and Morgan Stanley have pointed out that Baidu is China’s best artificial intelligence Internet company and a pioneer in promoting China’s artificial intelligence breakthroughs.
While leading industry breakthroughs, the fulcrum of AI native applications will also propel Baidu to soar into the sky and "generate" its own future.
Source: US Stock Research Institute