On September 7, the 2023 Tencent Global Digital Ecosystem Conference was held in Baoan, Shenzhen. Tang Daosheng, Senior Executive Vice President of Tencent Group and CEO of Cloud and Smart Industry Group, said that Tencent will enter the era of "fully embracing large models": "With large model generation technology as the core, artificial intelligence is becoming the key to the next round of digital development. Motivation also brings new ideas to solve industry pain points. Tencent Industrial Internet will continue to help industries strengthen through intelligence enhancement, data enhancement, and connection enhancement."
(Picture: Tang Daosheng, Senior Executive Vice President of Tencent Group and CEO of Tencent Cloud and Smart Industry Group)
At the conference, Tang Daosheng, senior executive vice president of Tencent Group and CEO of Tencent Cloud and Smart Industry Group, announced that Hunyuan, a universal large language model independently developed by Tencent, was officially unveiled to the industry.
Tencent's Hunyuan large model has a parameter scale of over 100 billion and a pre-training corpus of over 2 trillion tokens. It has strong Chinese understanding and creation capabilities, logical reasoning capabilities, and reliable task execution capabilities.
At present, the Hunyuan large model has been connected to more than 50 Tencent businesses and achieved initial results, including Tencent Cloud, Tencent Advertising, Tencent Games, Tencent Financial Technology, Tencent Conference, Tencent Documents, WeChat Search, QQ Browser and other businesses and product. Tang Daosheng said that thousands of industries can also call Hunyuan through API, or use Hunyuan as a base model to build large model applications for different industrial scenarios.
He also emphasized that Tencent has worked closely with 11,000 ecological partners and launched industry solutions covering more than 100 industrial scenarios to jointly serve thousands of industries. Last year, ecosystem partner revenue accounted for one-third of Tencent Cloud's overall revenue, and the revenue growth rate was four times that of Tencent Cloud. "On the road to exploring the future, Tencent is willing to join hands with partners to jointly explore new industry opportunities such as large models, continue to expand lower-tier markets, insist on polishing 'good products', embrace 'being integrated', and build a 'digital stage', please Partners should be the protagonists and jointly build the future of industrial intelligence."
The following is the full text of the speech:
Intelligence changes the future and data activates industries
Dear leaders, guests and media friends:
Good morning! Welcome to the Tencent Digital Ecology Conference. I am very happy to gather in Baoan, Shenzhen again. Here, I would also like to express my special thanks to the Shenzhen Municipal Government and the Baoan District Government for their strong support to us.
Although it has only been one year since the last partner conference, our cooperation with Baoan has progressed rapidly. Tencent Cloud Industrial Internet Headquarters Base has assisted the digital transformation of more than 300 Baoan companies; the "Greater Bay Area Artificial Intelligence Innovation Center" and the "Intelligent Network Technology Innovation Joint Laboratory" have been built to fully embrace the era of artificial intelligence.
This is the speed of Shenzhen racing against time, and it is also a true portrayal of how all parts of the country and all walks of life are accelerating to embrace the industrial Internet. Industrial digitalization is increasingly moving from idea to reality. Tencent's support for the real industry has become wider and wider, the scenarios have become deeper and deeper, and the capabilities have become stronger and stronger. Our positioning and collaboration with partners have become increasingly clear, and the division of labor has become increasingly clear, and the ecosystem has become increasingly prosperous.
We firmly believe that products are king, and we are determined to build solid technology and build a cloud product system with excellent performance, independent controllability.
Over the past five years, Tencent has invested more than 230 billion in R&D and achieved performance breakthroughs in many areas of core digital technology. For example, in the field of audio and video, Tencent Encoder won all 15 world championships in the MSU World Series at one time. For another example, by combining Tencent Cloud's high-performance Xinghai server, deeply optimized operating system Tencent OS, and self-developed database TDSQL, Tencent Cloud achieved a world record of supporting 814 million transactions per minute in the latest TPC-C performance test. record. In addition, we have obtained 33,000 patent authorizations worldwide in the fields of artificial intelligence, cloud native, audio and video communications, blockchain, and network security, with a total number of applications exceeding 66,000, ranking first among Chinese companies and second in the world. .
We focus on industrial scenarios, aim to solve pain points, promote the accelerated implementation of cutting-edge innovation, and help enterprises reduce costs and increase efficiency.
In the manufacturing field, Tencent's industrial Internet platform WeMake has been selected as a national-level dual-cross platform for three consecutive years, helping more than 600,000 manufacturing companies achieve better production management. In the field of travel, we have cooperated with 45 leading car companies to provide more than 8 million car owners with convenient and comfortable digital life in the car through Tencent Eco-connected Vehicles TAI. In the retail field, we use cloud and connection capabilities to help 600 leading retail companies improve sales performance and help brand development. More than 60 of them have sales exceeding 1 billion in the WeChat private domain. In the medical field, our AI guidance and diagnosis covers more than 1,000 hospitals across the country, providing patients with more accurate consultation and more detailed answers.
We let our partners play the leading role in building a high-quality new digital economy ecosystem.
We work closely with 11,000 ecological partners to launch industry solutions covering more than 100 industrial scenarios and jointly serve thousands of industries. We propose to focus on products, be a good provider of digital tools in the ecosystem, and leave more integration, delivery, sales and other links to ecological partners, asking partners to "play the leading role". They have clear positioning, clear division of labor, and complementary roles. Last year, ecosystem partner revenue accounted for one-third of Tencent Cloud's overall revenue, and the revenue growth rate was four times that of Tencent Cloud.
Tencent's champion products are widely adapted and integrated into partners' systems, becoming the "sharp knife" for partners to break through. For example, our database TDSQL is adapted to the business systems of more than 250 partners, covering key industries such as finance, government affairs, energy, and industry.
Based on the past and looking to the future, we are also facing new opportunities. As Mr. Ma Huateng, Chairman of the Board of Directors of Tencent, said, "The continuous enhancement of artificial intelligence technology has given enterprises an increasingly important production tool, which can continuously transform the 'information energy' hidden in data into high-quality development of the real economy. new momentum."
With large model generation technology as the core, artificial intelligence is becoming a key driving force for the next round of digital development, and has also brought new ideas to solve industry pain points: large language models make human-computer interaction experience simpler, understanding intentions and The ability to control the system is also more powerful, greatly improving the usability and ease of use of digitalization in the industry. The core center of the enterprise has also shifted from being driven by people to being driven by the collaboration of people and machines, greatly improving the quality of decision-making, the speed of response and the efficiency of execution.
Tencent Industrial Internet will also fully embrace the intelligent era. Through three "enhancements", we will help industries build "intelligent engines", tap the value of data, connect scenario applications, and achieve high-quality development.
First, we use "intelligent enhancement" to build "new momentum" for industrial development.
Today we are also very happy to tell you that Hunyuan, a general-purpose large language model independently developed by Tencent, has officially debuted for the industry!
Tencent's Hunyuan large model has a parameter scale of over 100 billion and a pre-training corpus of over 2 trillion tokens. It has strong Chinese understanding and creation capabilities, logical reasoning capabilities, and reliable task execution capabilities. At present, Hunyuan has carried out extensive docking and integration with Tencent's products, providing users with more considerate intelligent services in multiple business line applications such as Tencent Cloud, Tencent Advertising, Tencent Conference, Tencent Documents, WeChat Souyisou, etc. ,Improve work efficiency.
At the same time, we sincerely invite all walks of life to access the experience. Enterprises can call Hunyuan through API, or use Hunyuan as a base model to build large model applications for different industrial scenarios.
We always believe that large models need to be based on industrial scenarios and integrated with enterprise data to release the greatest value. AI helps improve industrial production efficiency and service innovation through its impact on all aspects of enterprise production, sales, and services.
Therefore, we launched the "Tencent Cloud MaaS service". In the Tencent Cloud industry large model selection store, we not only provide Hunyuan large models, but also list more than 20 of the latest and most popular open source general large models in the industry, as well as coverage Large industry models in more than 20 fields such as finance, cultural tourism, retail, government affairs, medical care, and education.
Enterprises can choose the appropriate model according to their needs, and then use Tencent Cloud Intelligence's TI platform to import the enterprise's proprietary data for further training and fine-tuning, and quickly generate more targeted and exclusive large models to meet the individual needs of the enterprise.
At the same time, we also combine large models with Tencent's industry solutions and enterprise-level applications to improve capabilities in production, marketing, after-sales services, office and other aspects, thereby reducing costs and increasing efficiency.
Through intelligent production, we solve the pain points of scenarios in innovative ways. For example, in the energy field, our country has a large number of power grid equipment scattered in remote areas. Relying on manual inspection methods, hidden faults cannot be discovered and dealt with in time, affecting the safety of power production. By training a large power vision model, one model can effectively adapt to different power grid scenarios. It has the advantages of high detection rate, low false detection rate, support for small sample training, and strong generalization ability, helping power grid production and operation, reducing burdens and increasing efficiency.
We improve service efficiency through intelligent marketing. In marketing analysis, Qidian Analysis·AI Assistant, based on industry models, allows sales staff to ask questions in natural language to obtain accurate business analysis. A PPT can be obtained in a few seconds, and there is no need to spend a lot of time learning. Complex software, making Kanban boards. The new generation of Tencent Qidian intelligent customer service is based on industry models and is trained and fine-tuned based on customer business needs. It can provide more accurate and detailed answers, and the resolution rate of complex problems has increased by 30% compared with before.
At Shanghai Renji Hospital, we combined digital intelligence with large-scale medical models and jointly developed the "digital intelligent medical customer service". Through this "virtual nurse", patients can use voice to provide barrier-free consultation to see which department they are in. Which doctor should I see and what tests should I do? The system's responses are also more accurate and user-friendly. This ensures efficient medical treatment for patients and avoids the waste of expert resources in tertiary hospitals.
At the same time, we also promote efficient collaboration through smart office. Now, applications such as Tencent Conference, Tencent Documents, Enterprise WeChat, and Tencent Enjoy have fully accessed large language model capabilities, helping to greatly improve work efficiency. For example, our corporate community product, Tencent Enjoy, has diverse capabilities such as training and learning, knowledge management, and cultural construction. During Tencent's internal use, it has accumulated 1.5 million technical and operational documents created by employees, with more than 50 million platform views per month. After Tencent Enjoy is equipped with large model capabilities, through AI assistants, it can help everyone quickly retrieve and refine information that is useful to them in complex databases. In digital training, AI can also answer course questions online and generate a test question bank with one click to facilitate more efficient corporate training.
Second, we use “data enhancement” to consolidate the foundation of enterprise digitalization.
Many people focus on large models, but not enough on data. Data has become a key production factor for society and enterprise operations. With the advent of the intelligent era, data will play a greater role and even become a core barrier to enterprise competition. In the era of "data explosion", solving problems such as data processing, mining, and security will become a mountain that enterprises and developers must climb.
How strong your data processing capabilities are determines how high your business development ceiling is. We continue to enhance our database, big data and other capabilities to meet the storage, reading, writing, calling and elastic management needs of massive data in enterprise operations.
Tencent Cloud Database TDSQL has a single instance storage scale of PB level and a single instance with over 10 billion daily requests. In the Garner report, its OLTP (online transaction processing) and lightweight TP capability scores ranked first in the country.
We also launched the AI-native vector database Tencent Cloud VectorDB, which is the first database in China that is fully AI-based from the access layer, computing layer, to storage layer. When you want to find a picture with "cat" in a large model, Tencent Cloud Vector Database can return query results from 1 billion vectorized pictures within 100 milliseconds.
Tencent Cloud Big Data helps enterprises realize big data application services such as data warehouse, user portraits, accurate recommendations, and risk management and control through different big data analysis engines and corresponding data development and algorithm development services. The computing power scale reaches tens of millions of cores, and the real-time calculation volume is hundreds of trillions per day, ranking first in the country in terms of scale and magnitude. Through the cloud-native data lake warehouse, we integrate the high performance and management capabilities of the data warehouse with the flexibility of the data lake; we achieve intelligent storage optimization, undifferentiated data access experience, and seamless connection between storage and computing engines.
In the digital era, the daily operations of enterprises are highly dependent on various digital systems. Data needs to be used in a safe and compliant manner. If a digital system is attacked and user data is leaked, it may cause significant losses to business operations and have a long-term negative impact on corporate reputation. Therefore, companies must attach great importance to and invest in the construction of digital security protection capabilities.
Tencent has industry-leading security capabilities. In areas such as threat intelligence, risk control and anti-fraud, data security, cloud security hosting services, and zero trust, Tencent Security has been rated as a "leader" and No. 1 in the industry by authoritative market analysis agencies for many years. position, it also serves more than 80% of financial industry customers, more than 90% of leading energy companies, and more than 80% of leading automobile companies.
Guangzhou Rural Commercial Bank used Tencent Security's "SOC+ Security Operation System" to intercept 20 million attacks with automated strategies. At the same time, daily threat analysis time has been reduced from the original 4 hours to 30 minutes, which effectively ensures system security and greatly improves security operation efficiency.
Third, we use "connection enhancement" to stimulate scene innovation.
To do a good job in industrial digitalization, we also need to promote data to be like water, constantly nourishing traditional production factors such as people, finance, and materials, so that a healthy and efficient enterprise management system can grow like a towering tree. Only flowing water has vitality, and flowing data can play its role. "Connection" is the "river" through which data keeps flowing.
In the past, we have connected more than 1 billion users through QQ, WeChat, etc. Today, we have also expanded this connection to the connection between people and people, people and things, and things and things in industrial scenarios. Enterprise WeChat connects more than 12 million real companies and organizations, and the number of Tencent Conference users exceeds 300 million.
In the industrial Internet, we are still committed to connecting users and integrating user identities across multiple enterprise applications. In the process, we have also discovered many pain points in enterprise identity ID management. In an enterprise's digital office environment, identity ID is the "work card" for employees to enter business scenarios. As there are more and more digital systems on corporate intranets and clouds, if different systems each have their own account "access control", the maintenance cost of multiple IDs will be very high, users will not be able to switch freely between different systems, and data will not be connected. , out of sync, which brings a lot of inconvenience to employees. At the same time, fragmented permission management also greatly increases the risk of data security.
Today, I am happy to announce to you that Tencent’s unified identity, ONE ID, is officially released. It can help enterprises uniformly manage employees' identity structures, login methods, and application permissions, so that they can be managed in one place and have consistent identities everywhere. Employees can smoothly access various business systems using one identity. One ID is not only compatible with self-developed SaaS applications such as Tencent Conference, Tencent Documents, Tencent Enjoy, etc., but will also be fully compatible with third-party applications. The unified identity capability of One ID also allows business systems to be freely formed into loose packages, allowing corporate customers to freely choose the applications that best meet their business needs and have a better business linkage experience.
In the era of true interconnection, we use real-time digital twins to connect time, space, objects and people, and integrate and process data based on technologies such as AI, real-time computing, simulation deduction, and big data to make the physical world observable and visible in the virtual world. Analyzable, predictable and controllable.
Tencent cooperated with the Forbidden City to create a "digital twin smart management platform" that not only replicated the appearance and structure of ancient buildings online, but also connected multiple management systems such as lighting, humidity, warehousing, and access control. It can automatically adjust parameters such as lighting, temperature and humidity, and carbon dioxide concentration according to different materials such as silk, ceramics, and bronze to provide the most suitable environment for cultural relics and minimize losses. At the same time, real-time monitoring and early warning are carried out for dangerous situations such as smoke, fire, and invasion to ensure the safety of cultural relics and buildings.
Dear guests and partners!
In the past five years, with the joint efforts of all sectors of society, digital technology has penetrated into all walks of life, and "going to the cloud" has changed from "consensus" to "reality." On the journey of the Industrial Internet, we have begun to climb the ladder of "intelligence" with all our strength and create a new situation in industrial upgrading.
On the road to exploring the future, we look forward to working with our partners to jointly explore new industry opportunities such as large models and continue to expand the lower market. We will insist on polishing "good products", embrace "being integrated", build a "digital stage", and ask partners to be the protagonists and sing a good show; with sustained and stable cooperation rules, we will make the division of labor between partners and Tencent clearer and more engaging. There is more accumulation and more predictable returns.
We will be based on the future and actual results, let digital tools be used well in the industry, and continue to help the industry strengthen with intelligence enhancement, data enhancement, and connection enhancement. We will continue to aim at creating value, firmly invest in technology and innovation, continuously improve products, enhance service experience, and work together with ecological partners and industrial customers to build an industrial intelligent future! thank you all!