laitimes

Robin Li revealed that the training speed of version 3.5 of the Wenxin model has been increased by 2 times, and the inference speed has been increased by 17 times

author:Red Star News

On June 26, Robin Li, founder, chairman and CEO of Baidu, attended the "Nishan Dialogue on Digital Civilization at the World Internet Conference" and delivered a speech entitled "Big Model Reshapes the Digital World".

The big model is the focus of global scientific and technological innovation and the main battlefield of the global artificial intelligence competition. Li Yanhong believes that "the key point of the new international competition strategy is not how many large models a country has, but how many native AI applications are on your large models, and to what extent these applications improve production efficiency." If we can squeeze into the table and get a ticket to the competition, China will have a stronger digital industry and the scale of the digital economy will grow tremendously. ”

Li Yanhong also revealed in his speech that Baidu Wenxin model has been iterated to version 3.5. Compared with version 3.0, the training speed is increased by 2 times, the inference speed is increased by 17 times, and the model effect is improved by more than 50%. "Version 3.5 of Wenxin model is not only a technical upgrade, but also a security upgrade." Robin Li emphasized, "Data quality, generation effect and content security have been significantly improved. ”

Li Yanhong believes that the digital economy driven by the big model as the key and the deep integration with the real economy will make the real economy stronger, better and bigger. In industries such as automobile manufacturing, energy, and transportation, large models can penetrate into core business scenarios and innovate in intelligent customer service, supply chain, system scheduling and other sectors, promoting the digital transformation and intelligent improvement of the industry.

The governance challenges brought by the big AI model cannot be ignored. Li Yanhong said, "Adhere to technological development and safe and controllable two-wheel drive, in order to achieve stability and long-term." If we navigate the path of AI safely and responsibly, big models will reshape the digital world, and AI can create unparalleled prosperity for the Chinese economy and the global economy, and improve the well-being of all mankind. ”

Red Star News reporter Hu Pei

Robin Li revealed that the training speed of version 3.5 of the Wenxin model has been increased by 2 times, and the inference speed has been increased by 17 times

The following is a transcript of the speech:

Distinguished leaders, distinguished guests, good morning!

I am very pleased to participate in the Nishan Dialogue on Digital Civilization at the World Internet Conference, and the theme of my speech is "Big Model Reshapes the Digital World".

In the past year, artificial intelligence has advanced at the iterative speed of "weeks" at all levels such as technology, products, and applications. The big model successfully compresses human cognition of the world and allows us to see the path to achieve general artificial intelligence. The next frontier of the development of large models is not only to imitate humans and complete human "prescribed actions", but also to help human beings to research and discover unknown areas and break through the limits that human beings have not broken through in the past. It would be even more meaningful if we could take that step.

How are big models reshaping the digital world? I would like to talk about two levels of technology and application:

At the technical level, in the era of artificial intelligence, the IT technology stack has undergone fundamental changes, from the original three-layer architecture of chip, operating system and application to a four-layer architecture of chip, framework, model and application:

The bottom layer is the chip layer, and the mainstream chip has changed from the CPU to the GPU. On top of the chip is the framework layer, and mainstream frameworks include Baidu Flypaddle, Meta's PyTorch, and Google's TensorFlow. Above the framework is the model layer, and ChatGPT and Wen Xin large models are at the model layer. Big models have become the operating system of the era of artificial intelligence, and all applications will be developed based on large models. Above the model is the application layer, which includes a wide variety of AI-native applications.

The structural changes in the IT technology stack mean that artificial intelligence, especially large model technology, will reconstruct the global digital industry. The key point of the new international competitive strategy is not how many large models a country has, but how many native AI applications are on your large models, and to what extent these applications improve production efficiency. If we can squeeze into the table and get a ticket to the competition, China will have a stronger digital industry and the scale of the digital economy will grow tremendously.

Baidu has been investing in artificial intelligence for more than 10 years, and has a full-stack layout in the four layers of chips, frameworks, models and applications, and has self-developed leading products and technologies in the four layers of architecture, so it can be optimized end-to-end and quickly improve the efficiency of large model training and inference. Wenxin big model is completely autonomous and controllable, we have achieved controllable data, controllable framework, and controllable model.

Of course, the governance challenges brought by the big model of artificial intelligence cannot be ignored. The application of new technologies often precedes norms, and the establishment and improvement of laws, regulations, institutional systems, and ethics to ensure the healthy development of artificial intelligence can create a good innovation ecology. Focusing on the future, while paying attention to risk prevention, we should also establish fault tolerance and correction mechanisms at the same time, and strive to achieve a dynamic balance between standardization and development.

Nowadays, large models have become hot. But 4 years ago, when the big model was not widely concerned, Baidu launched the Wenxin big model 1.0. Then continue to evolve to version 2.0 and 3.0.

Today, the Wenxin large model has been iterated to version 3.5, which has increased the training speed by 2 times, the inference speed has increased by 17 times, and the model effect has been improved by more than 50% compared with the 3.0 version in March.

Wenxin model version 3.5 is not only a technical upgrade, but also a security upgrade. We used the industry's mainstream large model basic capability assessment method to carry out the evaluation, and the results show that the 3.5 version of Wenxin large model has been significantly improved in terms of data quality, generation effect and content security.

The mainland artificial intelligence model has a certain foundation, and we need to catch up. At the same time, we should give full play to the advantages of application scenarios, further deepen vertical fields, build professional large models in finance, medical care, power and other fields, achieve technical optimization with high-quality applications and data feedback, help iteratively upgrade large models, and build a good AI ecosystem.

It is foreseeable that the big model will penetrate into more and more fields, and the digital economy driven by the big model as the key will be deeply integrated with the real economy, which will become stronger, better and bigger, create considerable incremental value, and bring profound changes in economic and social development and industry.

For example, in the automobile manufacturing industry, the most complex design link requires experienced engineers to find various combinations that meet the needs in more than 20,000 parts and hundreds of thousands of parameters, and then write documents and draw drawings. In Changan Automobile, large models can efficiently find combined information, automatically generate design documents, and greatly reduce the development cycle and cost. In Sinopec and China Southern Power Grid, large models can penetrate into core business scenarios, innovate in intelligent customer service, supply chain, system scheduling and other sectors, and promote the digital transformation and intelligent improvement of the industry.

In the field of transportation, intelligent transportation solutions supported by large models can improve the efficiency of traffic operation.

For example, on the last working day before the May Day holiday this year, Beijing's urban congestion index soared 2.5 times. From the second ring road to the sixth ring road, it is red, and the only green is Yizhuang. The traffic flow in Yizhuang has also increased significantly, but because of the deployment of the AI global information control solution, more than 300 intelligent intersections can automatically adjust traffic lights according to the traffic flow, and Yizhuang has become an "oasis" without traffic jams. The day before the Dragon Boat Festival, the traffic in Beijing and Yizhuang was strikingly similar: the city was congested, but Yizhuang was smooth.

In the Tai'an Taishan Scenic Area of Shandong, in order to serve the development of the tourism economy, promote the unblocking and smooth operation of urban areas, and solve the pain points of "parking difficulties" of foreign tourists, Baidu uses intelligent management and control methods such as traffic guidance screens and green belts to effectively ensure the safe travel of citizens and tourists.

Baidu's intelligent transportation solution has been adopted by 69 cities. By intelligently adjusting the timing of traffic lights, traffic efficiency can be increased by 15%-30%, which will drive GDP growth of 2.4%-4.8%.

In Jinan, Shandong, we have also landed Baidu Intelligent Cloud (Shandong) Artificial Intelligence Basic Data Industry Base, which not only cultivated new professions such as AI trainers, but also incubated 22 data annotation technology companies, driving regional employment and economic growth.

Whether from the perspective of technological trends or industrial applications, the big model is by no means a flash in the pan, but a major technological change that affects human development, an engine that drives global economic growth, and a major strategic opportunity that must not be missed.

Adhere to technological development and safe and controllable two-wheel drive, in order to achieve stability and long-term. If we navigate the path of AI safely and responsibly, big models will reshape the digital world, and AI can create unparalleled prosperity for the Chinese economy and the global economy, and improve the well-being of all mankind.

That's all for my remarks, thank you!