Author | Zhou Zhiyu

Editor | Zhang Xiaoling

Empowering everyone with programming abilities to create millions of 'super useful' applications.

At Baidu World Conference on November 12th, Baidu CEO Robin Li outlined a grand blueprint. By launching tools like Miao Da to Code, which greatly reduces the threshold for developing intelligent entities, the era of 'everyone is a programmer,' as mentioned by Li Yanhong earlier this year, is approaching.

Robin Li bluntly stated that this will be an unprecedented era where one can make money just with ideas.

Behind this grand vision is Robin Li's hope that Baidu's over a decade of AI technology accumulation can quickly realize commercial application landing with the trend of large models. Baidu's technology can also serve as the cornerstone in this transformative process, evolving into a vibrant ecosystem and driving productivity improvement through technology. In this process, Baidu can also find a stable 'second growth curve' beyond its advertising business.

To successfully take this step, landing AI applications is crucial. This is currently the key focus of Baidu.

Different from other asia vets companies talking about the prospects of AGI (General Artificial Intelligence) or Scaling Law in speeches, Li Yanhong was very practical at this year's Baidu World Conference, focusing mainly on asia vets entities, iRAG (Retrieval-Augmented Generative technology), and specific applications.

In the view of Li Yanhong, the biggest change in the industry in the past 24 months is that large models have basically dispelled illusions. In simple terms, when AI answers questions, the accuracy has greatly improved, the answers given are no longer 'serious nonsense,' but become usable and reliable.

This is thanks to the enhancement provided by RAG (Retrieval-Augmented Generative), which can further utilize retrieval information to guide the generation of text or answers, thereby improving the quality and accuracy of content.

From a segmented perspective, RAG's improvement on textual content is obvious, but there are still issues with details in multimodal content such as images, leading to illusions. Baidu focuses on iRAG technology, making images more realistic and eliminating the machine-like feel.

This is a commercially valuable scenario. For example, in the marketing of auto products, AI-generated images can maintain control over the details of auto products, improve product accuracy, and unleash creativity in the scene. This allows brands to significantly reduce costs in poster advertising while increasing authenticity.

Next, in scenarios such as movies, comics, comic books, etc., with the support of iRAG technology, applications like AI-generated images will further enhance production efficiency and reduce production costs.

The enhancement of basic model capabilities and technology lays the foundation for the explosive growth of applications.

At the conference, Li Yanhong showcased 100 major industrial applications based on large models, as well as multidimensional asia vets applications. The emergence of these applications is gradually ushering in the era of AI applications.

Robin Li has also contributed to the explosion of applications. He unveiled the 'Miaoda' tool at the conference, which helps people without programming background to develop applications by generating auxiliary code.

With the help of Miaoda, ordinary non-programmers can directly generate code under the assistance of large models, significantly reducing the development threshold. Combined with support for multi-agent collaboration and multi-tool mobilization, individuals can build a whole system through natural language interaction, thus improving efficiency.

This is what Robin Li said, that one can make money simply with ideas. In his view, developing intelligent agents is similar to creating websites in the PC era or managing self-media accounts in the mobile era. Intelligent agents may become the new carrier of content, information, and services in the era of AI natives.

Undeniably, the current AI industry seems to have entered a bottleneck period. The pace of model iterations by the AI industry leader OpenAI is not as expected; there is a continuous shift in the industry's outlook from unanimous optimism about whether AI model training can further skyrocket in performance. At the same time, with user growth slowing down, there has been no real 'killer app' emerging in the industry, signaling a decline in the AI hype.

Robin Li is optimistic. He believes that the slowing down of basic model iterations is a good thing for application development, and having a major version update every two years is a suitable pace.

He also does not believe that the absence of a 'killer app' at present is a bad thing. Citing examples like the steam engine revolution and the electrical utilities revolution, Robin Li compares AI to a new industrial revolution. Therefore, he does not simply compare it to the internet wave, expecting clear super applications to emerge after a few years of technology hype.

On the contrary, this will be a transformation akin to infrastructural change, a change in underlying technological capabilities. Similar to the philosophy of large companies like Alibaba, it is seen as being in the early stages of AGI transformation, with AI's greatest imagination lying in changing the physical world.

This will be an intriguing story. If Baidu's AI technology and applications can successfully evolve as part of the new era's infrastructure in this wave of AI, it can break free from the disappointment of the mobile internet era and find new growth points.

However, from the perspective of infrastructure transformation, this AI revolution will be a long process. At present, for major players with large models, they need to find directions that can be monetized, or even profitable, in order to gain investor approval.

China International Capital Corporation also pointed out that after the vigorous development and rapid iteration of large models in 2023-2024, the growth rate of total users has slowed down. Along with the gradual increase in the profitability of large model vendors in the reasoning process, the pace of large model vendors pursuing monetization is accelerating.

While the AI revolution brings future imaginative space, players like Baidu also need to continuously transform their businesses with AI, reflected in financial growth and ongoing commercial contribution. Even for ideal long-termists, they need to combine reality in order to turn technology into true value, supporting themselves in entering the new era of AI.