2023 Baidu Cloud Intelligence Conference: The intersection of technology and innovation

This
Baidu Cloud Intelligence Conference can be said to be full of highlights——

The newly upgraded Qianfan Large Model Platform 2.0, which includes 42 large models, 41 data sets, and 10 selected application paradigms, was released, the first large model ecological partner plan was released, and the release of Wenxin Large Model 4.0 was also announced. The model service report card is also very good: the number of monthly active enterprises has reached nearly 10,000, covering more than 400 scenarios in industries such as finance, manufacturing, energy, government affairs, and transportation.

And this conference made me feel brighter than Wenxin's big model, and the solution to reconstruct the four major industries based on the big model. It can be said that it has fundamentally completed the deconstruction of the industry.

As a Wenxin large-scale model with the largest industrial scale, its current industry breadth and coverage scenarios, as well as practical application experience, should also be well-deserved No. 1 in the country. As for my answer, I will simply continue with the part I am most interested in, and tell you about the solutions of these four major industries in detail, and fundamentally explore the way of "deconstruction".

Let’s first talk about the government solution for urban governance—Jiuzhou.

Kyushu, the name is the same as a novel of the same name that I like, and its pattern has really opened up. It is equivalent to reshaping a brain for the city. Its original core concept of digital government construction of "one-network unified management, one-network communication, and one-network collaboration" is really even more powerful for urban management.

For example, urban governance in the past was actually very complex, and generally required high-frequency scenarios and static analysis as the main governance methods. But now, with the help of a large model of one-network unified management, not only can we extract new thinking from the historical content of the city, understand problems and give the best suggestions in a timely manner, but also add more to the past governance methods in a sense. objective and scientifically based.

The "one-stop service" is mainly reflected in serving the people. The large model can understand all kinds of government affairs knowledge and terminology. Each citizen can have his own service assistant, which greatly reduces the labor force at the consultation desk and agency window.

One-network collaboration is mainly reflected in improving the efficiency of internal government offices. From official document writing to department coordination and decision-making support, office efficiency can be improved dozens of times.

What is the biggest benefit of using AI to improve things? It lies in objectivity and science, but it also lies in the execution of efficiency and the corresponding demand for the fastest delivery——

For regional governance like China, which has a huge land area and fragmented regions, this is simply a dimensionality-reducing blow. For example, grassroots urban and rural governance and local industrial applications based on different local conditions use AI to directly promote the integrated and coordinated development of urban and rural areas.

Let’s talk about the financial aspect, Baidu Intelligent Cloud Financial Solution: Kaiyuan.

In fact, the financial aspect is almost the original intention of many enterprises to develop large-scale models at present. The significance of many enterprises to create large-scale models is to serve finance. However, it is very difficult to achieve the scale of Wenxin large-scale model, from the technical threshold to the landing threshold.

Baidu's Kaiyuan has narrowed the problem from a macro perspective and focused on two levels to solve the biggest pain points in the financial industry -

Operational efficiency and business quality.

These two aspects are undoubtedly problems that the financial industry has always wanted to solve, but there is no particularly good way to solve them. The final solution is often only one: stacking manpower.

low efficiency? Adding people, poor business quality? Adding a elimination and replacement mechanism and a service quality evaluation mechanism will eventually require adding more people.

And through Baidu’s Kaiyuan, in terms of operational efficiency, through the compression and learning of massive financial professional knowledge, Kaiyuan has created a financial model that meets regulatory requirements for financial institutions, running through the front, middle and back offices of financial institutions’ business to achieve large-scale cost reduction Increase efficiency.

In terms of business quality, Baidu provides two sets of intelligent application solution suites: digital intelligent business operation and generative collaborative office. Support business growth.

If employees are liberated from the complicated and inefficient work of the past, will they still worry that the service is not good enough?
This is the perfect solution that can only be derived from the understanding and deconstruction of the industry at the bottom.

Let’s talk about the industrial aspect. Baidu’s naming really never disappoints me – Baidu Intelligent Cloud Industrial Solution Kaiwu.

In fact, our country has always been in a relatively awkward position with regard to the degree of digitalization of industry. The promotion of digitalization has indeed been ongoing, but the cost of actual implementation has always been high, which has greatly increased the survival costs of many small and medium-sized industrial enterprises.

And how did Kaiwu do it? Directly using the generalization capabilities of large industrial models helps industrial companies build more intelligent applications for production lines in less time with smaller data samples and computing power, and the large models’ own unique capabilities—

Generative automation capabilities can realize intelligent scheduling and management between factory-level models, use a large language model to connect intelligent applications in different production lines, and achieve global resource allocation and intelligent data flow at the enterprise level or even the industry chain level. The seamless transmission of data will naturally drive up the efficiency of the industry.

Finally, let’s talk about the intelligent transportation solution: ACE3.0.

Baidu has really put a lot of effort into urban transportation.

In the past five years or so, Baidu has made many achievements in the field of intelligent transportation, and has become the only company in the world that has realized the closed-loop core technology of the car, road, cloud, and map full-stack. In this process, Baidu is Accumulate to the greatest extent based on the basic elements of transportation, such as dynamic and static traffic data, transportation industry knowledge map and engineering capabilities.

Utilizing the powerful knowledge compression capabilities of large models, ACE 3.0 can aggregate and generalize the basic capabilities of vehicles, roads, clouds, maps, and all transportation elements to develop intelligent transportation solutions based on regional specificities.

For example, for the traffic police department, there were originally only dozens of congestion identification factors, but now they can be increased to one or two hundred through large models, and the traffic organization plan can be redesigned faster and better based on the large model. And optimize, and keep expanding this range.

The above is the interpretation of industry solutions at this Baidu Cloud Intelligence Conference. These solutions are outstanding. They not only cover the B-side and C-side business scenarios, but also effectively realize China Unicom. It may be that the actual implementation value has not been fully released yet, but when the technology With more maturity and more stable computing power, a new migration in the Internet era may come earlier.

I also look forward to the official arrival of an era of large model problem solving.

​