In the era of large-scale model applications, Baidu made a head start

Original link:

With the first batch of large-scale model applications going online through filing at the end of August, China’s artificial intelligence large-scale model market has entered a new stage. The large-scale model products developed by technology companies and institutions that have passed the filing can provide services to all users, while before it could only be carried out by the number of people. Limited testing.

Many large-scale model practitioners believe that there will be a large number of large-scale model applications in China. However, there are still many unknowns about how to use the ability of large models to solve the common needs of a large number of users to create popular apps. In the past half a year of exploration, some fast-moving large-scale companies and entrepreneurs are trying to provide references.

AI-native applications are the meaning of the existence of large models

On September 5th, Baidu’s “Wenxin Cup” Entrepreneurship Competition, which lasted for more than three months, ended, and 15 teams were selected as winners. In addition to receiving tens of millions of yuan in investment from Baidu, they will also invest in technology, products, and development strategies. , Capital cooperation and other aspects have obtained Baidu’s long-term support.

Baidu CEO Robin Li said in his award speech that a good native AI application should meet at least three conditions: support natural language interaction, support understanding, generation, reasoning, and memory information, etc., and the interaction should not exceed two levels of menus. Applications should be able to solve problems that could not be solved or not solved well in the past, rather than simply repeating mobile Internet applications or computer software.

“The model itself does not directly generate value, and the application developed based on the basic large model is the meaning of the model.” Li Yanhong said that the operating systems in the mobile Internet era are only Android and iOS, but there are many particularly successful applications. In this case, the basic large model is the operating system, and the new functions implemented based on the large model are “native applications” in the era of artificial intelligence.

In the new pattern of artificial intelligence, the ability of large models is the foundation, which directly determines the upper limit of the application of large models, but large models alone cannot form a prosperous ecology, and applications based on large models are also the key.

Previously, China’s large models were mainly model development, and the application layer was often ignored. But if everyone only focuses on model development and no one develops applications, it will be like a car without wheels. Overseas, the large model application layer has already begun to develop. In China, Baidu is a start.

“We hope that entrepreneurs can create explosive applications in the AI ​​​​era based on the Wenxin model.” Robin Li said. Baidu’s entrepreneurial competition that started in the stage of large-scale model testing has now come to an end, but there is no sign that Baidu’s actions to cultivate a large-scale model entrepreneurial ecology will stop.

The 15 winning apps cover multiple segments

Document-based question-and-answer assistant, design creative assistance, medical content generation, new material discovery, two-dimensional content creation, salesperson training… The projects of the winning teams of the Baidu Big Model Entrepreneurship Competition cover multiple subdivisions. Together, they exhibit many of the characteristics of large-scale applied entrepreneurship.

First, entrepreneurs working on the application of large models usually have a deep understanding of the scenarios they are trying to change. It is precisely because of the in-depth understanding of specific application scenarios that entrepreneurs can better understand the problems faced by existing solutions in these scenarios, and thus are more likely to use large models to develop better solutions.

Paoding Technology, the ChatDOC company that won the first prize in this competition, was founded in 2017. In the six years since its establishment, it has been working on products related to financial documents, such as writing prospectuses based on a large amount of information provided by customers. The leader of the ChatPPT project, which also won the award, Zhou Ze’an has 10 years of experience in PPT engine and function development. The previous entrepreneurial project Pocket Animation was acquired by WPS.

Large-scale startups that continue to gain customers and investment overseas are similar. For example, Jasper AI, a large-scale model application company with a valuation of 1.5 billion, mainly uses the capabilities of large-scale models to provide marketing support for customers. Its founding team has a deep marketing background. In November last year, OpenAI led a $27 million investment in language learning application Speak. The founder made an application to assist learning and memory in high school and sold it.

Secondly, the entrepreneur’s strategy is not to build a new solution based on the large model, but to use the large model to optimize a specific link of the original solution. In this optimization process, they maximized the unique advantages of the large model while avoiding its remaining weaknesses.

“Our core is not the big language model, but how to make an artificial intelligence capable of playing the role of a human shopping guide.” Chen Lifei, founder of Buysmart.AI, said that the root lies in how to make it correctly understand the user’s problem, and then recommend the products you want . Buysmart.AI is another project that won the first prize in the Baidu Big Model Entrepreneurship Competition. The approach they take is to combine large models with recommendation algorithms to exploit their ability to understand and process large amounts of information.

Lin Demiao, CEO of ChatDOC, said that similar smart document products on the market often answer questions that are not asked and generate answers by themselves. Therefore, they will limit the ability to generate large models in ChatDOC, and require each answer it gives to quote the original text. If no suitable original text is found, they will feedback “not found” instead of random answers.

In Magic Quantity, which uses large models to assist in the discovery of new materials, the greatest value of large models is to assist in the construction of a computing experiment platform that can call advanced algorithms and experiments without mastering the code. “The large language model reduces the cost of using software or operating each instrument to a certain extent, and can directly implement specific operations through voice.” Liu Yuyang, founder and CEO of Magic Quantum Technology, said.

Third, for large model applications, although it is not difficult to switch the underlying large model, if the same large model is used for a long time, dependencies may arise. Many of the teams participating in the Baidu Large-scale Model Entrepreneurship Competition this time developed applications overseas based on ChatGPT in the early stage, while their domestic business has now switched to Wenxinyiyan.

“It is not as difficult as expected to replace an overseas model with a domestic model. After we change, we don’t need to make too many changes, and the whole process can run.” said Chen Lifei, founder of Buysmart.AI.

Peng Kangwei, CEO of Genie AI, which uses large models to assist in the creation of two-dimensional content, also has a similar feeling. Switching from ChatGPT to Wenxinyiyan, “The fine-tuning of the model and the expression of some keywords are not much different. From safety and compatibility From the perspective of Chinese, Wen Xin Yi Yan would be better.”

However, in her view, if you use a large model such as ChatGPT for a long time, if you want to fully utilize its capabilities, you must set up a product architecture and build code around it. “Switching models after a long time is costly.”

This is an often overlooked aspect of the big model competition. Developers of large-scale model applications may rely on a specific large-scale model, which means that those large-scale model suppliers that enter the market earlier and attract entrepreneurs earlier will have greater advantages.

Large models gradually enter the era of native AI applications

Before the implementation of the large-scale model policy, most companies were quite cautious in promoting the application of large-scale models. Their products for individual users are usually in the stage of internal testing or invitation testing. Ordinary users cannot directly register or use them, and companies do not actively advertise to promote large-scale products. These factors limit the speed of product dissemination.

The new environment created by the implementation of the policy has transformed the competition of large models into a contest of comprehensive capabilities: the key to success is no longer just a company’s technical strength in training large models, but also its insight into market needs, development of matching applications, and excellent operating capabilities.

This is a test for every startup that develops large-scale applications. For companies developing fundamental large models, the test is also directly related to their ability to build ecosystems. This may be a direct reflection of their competitiveness.

Li Yanhong believes that “only the best large-scale models can grow the best artificial intelligence native applications.” He said that Baidu will soon launch version 4.0 of the Wenxin Large Model, with the goal of “Baidu’s goal is to build a good basic capability for large models and support the development of artificial intelligence native applications.”

Baidu’s investment in large-scale model ecology is also continuing. It is understood that, in addition to the “Wenxin Cup” entrepreneurial competition, Baidu also launched AI Studio Galaxy large-scale model community, plug-in mechanism and Wenxin large-scale model “Galaxy” co-creation plan for developers, attracting more people to join Baidu’s large-scale model ecology .

According to the data disclosed by Baidu, nearly 10,000 enterprises are currently active on the Baidu Smart Cloud Qianfan large-scale model platform every month, covering more than 400 business scenarios in industries such as finance, manufacturing, energy, government affairs, and transportation.

At the end of August, Baidu’s Wenxin Yiyan was officially opened to the public after being filed with relevant departments. According to Baidu, Wenxinyiyan answered 33.42 million questions from netizens on the first day of its opening. “A large amount of real human feedback will help Baidu improve the basic model quickly and efficiently.”

“I believe that the Wenxin large model will become the first choice for AI entrepreneurs and developers. More and more applications will be built on the model, and the entire ecosystem will be full of vitality.” Robin Li said.

Title map source: Baidu

This article is transferred from:
This site is only for collection, and the copyright belongs to the original author.