China's "black technology" competed in the Paris Olympics, and the application of AI large models has become the focus of competition

Author | "Caijing" new media writer Shu Zhijuan/text editor | Wang Jingya

In the early morning of July 27, Beijing time, the world entered the Paris Olympics to see the grand occasion of the event. In this Olympic Games, China's black technology will shine. The author was informed before the competition that Chinese technology companies, including Alibaba Cloud, SenseTime, HighGreat and Absen, will bring their cutting-edge "black technology" to the Olympic stage.

From the 500 times per second recognition of the chip football, the sports mat equipped with smart chips, the dazzling LED floor screen, to the Olympic fireworks in the night sky of Paris and the wonderful performance of 1,100 drone formations, to the cloud computing replacing satellites as the main way to broadcast the Olympics, the application of robot attendants and unmanned vehicles...... The ability to innovate in science and technology is demonstrated everywhere.

Since the beginning of this year, with the continuous heating up of the "100-model war", the focus of competition between enterprises has shifted from the simple competition of technology, data and computing power to the application level. From a global perspective, building a large model ecosystem has become the key to future development. However, as an emerging technology, the application of AI large models is still in its infancy, and the construction of the ecosystem faces multiple challenges. From the perspective of the industry, the Olympic Games, as an international sports event, is undoubtedly a "training ground" for the construction of an AI large model ecosystem, and the application of large models will be accelerated through cross-industry and cross-border cooperation.

Cloud broadcasting has become the mainstream method of content broadcasting

When the Olympic flame was lit on the Seine, China's "Olympic Team" joined hands with Chinese technology to demonstrate the innovative power of "artificial intelligence+" to the world.

As an indispensable part of the Olympic Games, the technical innovation of the broadcast of the event is particularly eye-catching. According to the Paris Olympic Organizing Committee, cloud computing has replaced satellites as the main way to broadcast Olympic events in this Olympic Games. This innovative initiative means that more than 11,000 hours of live content will be seamlessly delivered to billions of viewers in more than 200 countries and regions around the world with the help of Alibaba Cloud technology, an increase of nearly half of the live broadcast time compared to the Tokyo Olympics.

In the past, live Olympics relied heavily on satellites and traditional fiber optic cables, which were costly and limited in flexibility, often leaving broadcasters helpless in the face of unforeseen situations. Today, the introduction of cloud computing has changed that completely. Not only does it dramatically cut broadcast costs, but it also delivers a richer live streaming experience with lower latency. What's more, broadcasters are freed from heavy hardware equipment constraints and can achieve flexible and efficient on-site deployment, which greatly optimizes resource allocation.

This technological leap is not unfounded, as cloud technology has been on the rise since the Tokyo 2021 Olympic Games. According to Sotiris· Salamuris, chief technology officer of the Olympic Broadcasting Service (OBS), at the time, a broadcaster arrived at the end of the day, but they had very demanding requirements to get the entire live content and broadcast it to their home country in UHD format. At this late juncture, this requirement is a huge challenge. As a result, OBS tried to use the cloud relay solution for the first time. "I was a little nervous, not sure if the attempt would succeed, and the result was far more than expected." So said Sotiriss · Salamulis.

"Cloud broadcasting can bring some movie special effects-style effects into the broadcast of the event, such as watching the game from a high-degree perspective similar to 'bullet time'." The relevant person in charge of Alibaba Cloud told the author that with the help of the high concurrency and low latency characteristics of the cloud, dozens of high-definition images are transmitted in real time, and then millisecond-level synthesis and modeling are carried out by intelligent algorithms, bringing ultra-slow-motion and panoramic viewing of movie special effects into the broadcast of the Paris Olympic Games.

It is reported that Alibaba Cloud has deployed AI enhancement technology in 14 venues and launched a multi-shot playback system to achieve "space-time freeze", which can freeze the continuous actions of the timing sequence into 3D space. This technology not only allows the audience to clearly capture every detail of the athlete's movements, but also plays back through 360-degree free perspective, so that the wonderful moments can be presented in an all-round and multi-angle way. As a result, OBS can achieve a high degree of freedom playback in the live broadcast signal that can only be generated in the past 10 hours.

In the view of Liu Xingliang, president of the DCCI Internet Research Institute, during the Paris Olympic Games, the audience can freely choose the perspective of 360 ° to watch the game, and the game picture can be viewed from any angle, which is the audio-visual effect brought by the combination of high technology and live broadcast of the game. However, this also puts forward higher requirements for cloud computing security, data storage and management capabilities, and related technologies and processes of cloud vendors.

Large AI models improve training efficiency

At this year's Paris Olympics, the application of AI large model technology has become a major highlight. Many teams of the Chinese Olympic delegation have made use of "AI+" high-tech equipment to optimize their preparation training and competitive performance. From generative AI models to intelligent image recognition to virtual personal assistants, these cutting-edge technologies are being applied to all aspects of pre-competition training to tailor training plans for athletes and accurately control their competitive status.

The sports model jointly developed by Baidu and Shanghai Sports University provides important support for the daily training of many national teams such as diving, swimming, and track and field, as well as the preparation for the Paris Olympic Games. Through cooperation with Lenovo, the Chinese national sailing and windsurfing team has introduced a laptop with a built-in personal agent "Xiaotian" to assist coaches in formulating personalized training plans, and can also carefully analyze athletes' body movements and movement trajectories, providing unprecedented training insights for the coaching staff. The Chinese national basketball team used the SenseTime model for data statistics and game video processing.

In addition to providing training and game analysis for athletes, AI large model technology also plays an important role in event commentary. The addition of the Tongyi Qianwen model has brought both professionalism and accuracy to the official commentators of the IOCOG, so that the global audience can enjoy a more exciting interpretation of the event.

In the Olympic Village, the Chinese robot master of the smart restaurant has become the focus. Not only are they able to efficiently prepare delicacies such as xiao long bao and siu mai, but they also have the ability to judge satisfaction through customer expressions, enabling instant service optimization. In this Olympic Games, various robots developed by China are located in more than 20 venues, from catering and cleaning to security and medical care, showing the infinite possibilities of intelligent services in an all-round way.

In terms of hardware, Taishan Sports' intelligent sports mat has also attracted everyone's attention. For the first time, the sports mat integrates intelligent chip technology, which realizes the traceability of the whole life cycle status of the equipment, and can accurately capture the strength, speed and movement trajectory data of athletes, providing data support for training and competition in judo, wrestling and other events.

Song Xiaogang, executive director and secretary general of the China Robot Industry Alliance, pointed out that with the rapid development and in-depth application of AI large models, big data and other technologies, intelligent development is entering a new stage of multi-dimensional upgrading. At present, the industry is in a period of rapid growth, attracting a large influx of capital, and the popularity of the industry continues to rise. The application of these technologies in the Paris Olympics not only greatly improved the scientific and technological content of the event, but also provided practical experience for the future development of intelligence.

The ecosystem is the key to the competition of cloud vendors

The frequent appearance of China's hard-core technology at the Olympic Games reflects to a certain extent the core competitiveness behind it - the ability to build an ecosystem.

A cloud vendor who did not want to be named told me that the success of AI models does not depend solely on technical performance, but also on whether they can be seamlessly integrated into a comprehensive and powerful ecosystem. This ecosystem, like a sophisticated machine, covers every key link from data collection, model training, inference deployment to application development and user feedback, and its efficient operation directly determines the overall performance of the large model and the quality of user experience.

However, as an emerging technology, the construction of its ecosystem is still in its infancy and faces many challenges. On the one hand, technology is changing rapidly, and enterprises need to constantly track the latest trends to ensure that ecosystem technologies stay ahead of the curve and adapt to the uncertainty of the technology environment. On the other hand, the diversity and variability of user needs also require the ecosystem to be flexible and scalable enough to adapt to the changing needs.

The relevant person in charge of Alibaba Cloud said that through the Olympic Games, cross-industry and cross-border cooperation has been carried out to create customized solutions and apply large model technology to more practical scenarios, which not only expands the application scope of large models, but also injects vitality and innovation into the ecosystem.

Baidu Intelligent Cloud pointed out that whether the large model can take root depends on whether it can be combined with the application scenarios of thousands of industries, and international sports events such as the Olympic Games provide excellent application scenarios for the large model.

In Liu Xingliang's view, it is expected that by 2024, the diversified development of AI large models will continue, and the domestic competition pattern has not yet stabilized. In this context, in addition to competing in the price war, domestic large-scale model enterprises should focus more on the construction of the ecosystem. This includes improving the industrial chain, enhancing technological innovation capabilities, and maturing the innovation ecosystem, while strengthening data security and privacy protection to achieve breakthroughs in the fierce market competition.