Written by Chen Dengxin
Editor / Li Ji
排版/ Annalee
The large model has landed, and there has been new progress.
On April 16, 2024, the 2024 Baidu Create AI Developer Conference released the Wenxin Large Model 4.0 Tool Edition, the world's first AI native operating system, DuerOS X, etc., among which the "out-of-the-box" AgentBuilder, AppBuilder and ModelBuilder are particularly eye-catching.
This means that the threshold for the development of AI-native applications has been lowered.
So, today, why is the outside world gradually reflecting on the large model? "Everyone is a developer", is it the winner or loser of the landing of the large model? The "last mile" of the large model is from business to ecology?
Give up the thinking of explosive styles, "Internet celebrities" can become "long-term red"
The attitude of the outside world towards the large model has changed subtly.
In foreign countries, OpenAI encountered the problem of sluggish user growth, and had to allow unregistered users to directly access the free version of GPT-3.5 in order to increase its attractiveness and convert more large model users.
According to Similarweb, ChatGPT received 1.77 billion visits worldwide in March 2024, down from a record of 1.81 billion in May 2023.
ChatGPT降温的势态,肉眼可见。
In China, large models continue to "involute", volume computing resources, volume parameters, volume application scenarios, volume commercial marketing...... But the actual experience of the public has not changed much.
In other words, the phenomenon of "applauding but not applauding" has appeared in large models.
In March, the top list of domestic web AI products, source: silicon-based position
In this regard, at the 2024 Baidu Create AI Developer Conference, the industry has an answer to get out of the predicament: the landing of large models should not uphold the "explosive thinking", but seek "equal rights in AI development", and promote the whole chain of "research→production→ supply→ sales→ and services" to reduce costs and increase efficiency.
In layman's terms, the large model must exist like "oxygen" in order to change from "Internet celebrity" to "ever-popular".
Robin Li, Founder, Chairman and CEO of Baidu, said: "Large models themselves do not directly create value, and AI applications developed based on large models can meet real market needs. ”
At the scene, Robin Li shared specific ideas and tools for developing AI native applications based on large models, "This is what our Baidu has stepped on countless pits and paid high tuition fees in exchange for the practice of the past year."
Historical experience shows that any disruptive technology from birth to popularization, will not be achieved overnight, there is no lack of aggressiveness and impatience, which puts forward higher requirements for the strategic height and strategic determination of enterprises.
An Internet observer told Zinc Scale: "Because ChatGPT, Sora, Kimi, etc. have become popular successively, the voice of teaching Baidu to do things is not uncommon, but Baidu is not optimistic about 'entertainment first' at the beginning, but recognizes that 'efficiency is king', and takes the route of completely taking root in the industry, and strives to solve the practical problems of the industry, after all, only when the industry crosses the 'gap' of technology, can the large model prove its commercial value, ChatGPT, Sora, Kimi's high and low voice confirms this. ”
From this point of view, Baidu has been doing the hard but right thing.
AI development is equal, so that there is no hard-to-write code in the world
It can be seen from the above that AI development equality is the key to the implementation of large models.
"In the past, developers used code to change the world, but in the future, natural language will become the universal programming language. Robin Li said that AI is setting off a creative revolution, "In the future, developing applications will be as simple as shooting short videos, everyone is a developer, and everyone is a creator." ”
In order to make the world have no hard-to-write code, Baidu has launched three major AI development artifacts.
First, zero-code agents are created.
In the past, there was no way to develop applications without knowing code, but with AgentBuilder, an agent building platform based on the Wenxin model, you can break this common sense, and use the zero-code or low-code development model to easily create 7X24 hours online gold salesmen, education consultants, etc.
Ma Baoyun, head of Wenxin Intelligent Twins Ecology of Baidu's Mobile Ecosystem Business Group, said: "AgentBuilder's toolbox is like a stationery box and a jewelry box, each of which can be directly accessed, and developers no longer need to write code and develop a little bit by themselves. ”
In the first week of launch, it successfully distributed 1.55 million times and interacted with users 58,000 times, resulting in a linear increase in the number of lead conversions, a significant reduction in the conversion cost of effective leads, and a significant improvement in operational efficiency.
Second, develop AI-native applications with zero threshold.
If you want to develop more complex programs, you can use AppBuilder, which encapsulates and presets the various components and frameworks needed to develop AI-native applications.
AppBuilder includes advanced components for typical application scenarios, such as RAG for quizzes, code interpreter with computing power, and generative data analysis GBI, as well as a variety of basic components, and supports the integration and extension of third-party components, making it easy for developers to flexibly assemble.
"In just three steps, developers can develop an AI-native application in natural language, which can be easily published and integrated into a variety of business environments. Robin Li said.
AppBuilder architecture
Again, fine-tune the model from scratch.
ModelBuilder provides 77 mainstream large models at home and abroad, including the Wenxin large model series, which is the development platform with the largest number of large models in China, and supports one-stop model fine-tuning.
In short, ModelBuilder can customize models of any size according to the needs of developers, and further fine-tune the models according to the subdivision scenario, so as to find a balance between effect and cost.
According to public information, ModelBuilder has served 85,000 enterprise customers, fine-tuned more than 14,000 models, and provided support for the development of 190,000 AI-native applications.
It is not difficult to see that the development requirements covered by AgentBuilder, AppBuilder and ModelBuilder are from easy to difficult, and the same is natural language development.
In this way, thousands of AI-native applications can be incubated to solve a variety of real-world problems.
Ecological prosperity is the ultimate destination of the big model
In addition to AI development tools, Baidu also launched the second "Wenxin Cup" entrepreneurship competition.
The second "Wenxin Cup" has a wider scope of project screening, facing the global market and college students, helping entrepreneurs and developers to develop various AI native applications, and setting up a "special award" for the first time, with up to 50 million cash and resource support.
It can be seen that the prosperity of the AI ecosystem has become Baidu's core demand.
After all, with the blessing of Wenxin Model 4.0 and tools, the development threshold of AI native applications has dropped sharply, and without worries about computing power, training, machine learning, etc., you can focus on meeting user needs, so the entry of individual developers or small and medium-sized startups will no longer be unattainable.
In fact, Baidu has explored a new path of large models: in the era of large models, Baidu seeks to "empower everything", and is willing to be an AI empowerment platform to achieve a win-win situation with partners.
Developers and entrepreneurs have limited energy and resources, and it is more efficient to focus on applications without having to make both models and applications, while Baidu can also continue to expand its base by actively supporting, incubating, and encouraging third-party AI native applications.
In this way, the gift of roses leaves a lingering fragrance.
Wenxin Yiyan transcript
A market source told Zinc Scale: "If you regard large models as a business, you will consider more short-term interests such as volume, traffic, and sales, and it will also cause developers and entrepreneurs to worry about jobs." ”
The above-mentioned market participants further said that from business to ecology, long-term interests will become the "baton", and the large model is bound to become the foundation of Baidu's sustainable development, which not only reshapes the new ecology of the Internet, but also can recreate "Baidu".
Robin Li said: "In the mobile era, WeChat has not eaten Pinduoduo, Didi has not become a part of Tencent, they all provide their own unique value, have their own very different competitiveness, their rise is dependent on a closed platform in the mobile ecosystem - WeChat, but they are not afraid of WeChat to grab its job, so there is no need to worry about the application of AI in the basic model." ”
It can be said that Baidu is at the forefront of the times in terms of its open and inclusive attitude towards developers and entrepreneurs, or its original intention of building China's large model ecosystem, and once its large model ecosystem moves from embryonic to mature, it is bound to become the new standard-bearer of the AI 2.0 era.
There is no doubt that Baidu is blazing a new trail.