AI big model, Alibaba is bound to win.
Alibaba CEO Wu Yongming made his first public speech.
"In the AI era, Alibaba will become an open technology platform enterprise that serves the AI innovation of the whole society."Wu Yongming, CEO of Alibaba Group, summed up the new orientation and direction of Ali in the future.
A few days ago, Wu Yongming boarded the podium of the plenary session of the 2023 world internet conference Wuzhen Summit. This is the first time that Wu Yongming has delivered a speech at a public event since he became CEO of Ali Group.
Wu Yongming emphasized the importance of AI computing in his speech. "The AI-driven cloud computing system is surpassing the traditional computing system and becoming the cornerstone of the digital world, and will eventually take over all computing resources."
Source: Alibaba WeChat official account
For the future of the AI era, Wu Yongming predicted that life products and corporate work will become more intelligent. "Every enterprise will also be equipped with AI assistants. Just like our smart cars today, assisted driving and autonomous driving have become standard."In his view, the era of intelligence has just begun, "everything is possible."
AI has become the key to the innovation breakthrough of digital economy in China. Wu Yongming positioned Ali as a "science and technology platform enterprise". This strategic decision not only conforms to the development trend of the times, but also meets the needs of the enterprise’s own development and upgrading.
In the past 20 years, Ali has built digital business basic services covering transactions, logistics, payment and production, which has provided strong support for the development of the e-commerce industry.
In the AI era, Ali should take the route of "technology platform enterprise" to provide infrastructure for AI innovation and transformation of enterprises in the whole society. Ali will open a set of technical systems and infrastructure capabilities to the whole society in the form of platform services, and extend the platform capabilities from commodity trading to technology.
Not every enterprise has the ability to develop large-scale AI models. And Ali is helping companies solve this problem.
It is understood that Ali founded and operated the China AI model open source community "Magic Ride" last year. Since the "Magic Ride" community was launched one year ago, the model download volume has exceeded 100 million.
Not long ago, Ali also released a one-stop big model service platform-Alibaba Cloud "Bailian", which helps small and medium-sized enterprises and developers to carry out AI innovation with lower cost and higher efficiency, further lowering the threshold for AI model development.
At the "Internet Entrepreneur Forum" of Wuzhen Summit, Wu Yongming revealed that Alibaba is about to open a large model with 72 billion parameters, which will be the largest open source model in China.
Alibaba Cloud has successively opened up the 14 billion parameter model Qwen-14B and the 7 billion parameter model Qwen-7B. At present, these two large models have been launched in the ModelScope magic community and can be commercialized for free.
Qwen-14B Source: ModelScope Magic Community Screenshot
Wu Yongming’s clarification of Ali’s position and direction as a "technology platform enterprise" means that Ali already has perfect infrastructure construction technology in the big model era, and is ready to achieve a win-win situation with all social enterprises.
In order to serve more enterprises and AI developers, Alibaba will insist on doing two things: First, provide a stable and efficient AI basic service system, especially strong cloud computing capabilities, and build a solid foundation for training AI in the whole industry and using AI in the whole society; The second is to create an open and prosperous AI ecosystem.
Wu Yongming believes that "AI+ cloud computing" two-wheel drive is Alibaba Cloud’s underlying ability to face the future. Alibaba will base itself on the position of "technology platform enterprise" and build a more solid infrastructure base and a more open and prosperous AI ecosystem.
AI drive is one of the strategic focuses of Ali.
On September 10th, Wu Yongming officially became the CEO of Alibaba Group. On the 12th, Wu Yongming issued a letter to all employees, announcing the establishment of two strategic focuses: customer first and AI drive.
In Wu Yongming’s view,"In the next decade, the biggest variable is undoubtedly the profound changes in the whole industry brought by AI. We must reshape the user value with the determination to start a business again in the AI era, so that we can get the opportunity to continue serving customers in the next decade. "
At the same time, the application and landing of AI technology need to be user-oriented. Wu Yongming hopes to turn a large number of user scenarios of the group’s businesses into application fields of AI technology, bringing a breakthrough user experience and business model.
Ali’s Taotian Group has recently launched a series of AI tools. For example, intelligent generation of model drawings, official customer service robots, Wanxiangtai Unbounded Edition, Ali Mama Bailing and other businesses operate AI tools.
Official customer service robot page Source: Taobao
During the preparation and sale period of double 11 this year, merchants on Taobao Tmall platform have called AI tools more than 1.5 billion times.
Taobao also launched the consumer-oriented AI intelligent assistant "Taobao Ask" this year. According to statistics, Taobao has been open for testing for less than two months, and the number of invited users has reached 5 million. On average, highly active users interact with Taobao for 8 times a day, and some users even ask questions for 4,000 times.
In terms of globalization, Alibaba International Station has launched an AI business assistant to help foreign trade merchants in China go to sea efficiently. The intelligence of Ali’s various businesses has further deepened.
The development of Ali AI technology is inseparable from the infrastructure construction in Alibaba Cloud. Zhou Jingren, CTO of Alibaba Cloud, said earlier that Alibaba Cloud has built a perfect AI infrastructure and a perfect AI big model training platform, which can provide you with a very perfect model reasoning service.
It is understood that Alibaba Cloud recently officially released the 100 billion-level parameter model Tongyi Qianwen 2.0. At present, more than 30 cloud products in Alibaba Cloud have been connected to the big model capability.
Source: WeChat official account, Alibaba Cloud
The construction of large models often requires a lot of computing power and consumes manpower, material resources and funds. If enterprises build their own large models, the cost problem will become more and more serious. Li Yanhong, founder, chairman and CEO of Baidu, believes that "the best solution is to integrate the ecology of large models with plug-ins or APIs and create solutions for their own industries."
Alibaba Cloud has perfected the AI infrastructure in order to help all walks of life use AI technology to solve practical problems and focus on application development without considering the cost.
Zhou Jingren once said, "At present, half of the large model enterprises in China are running in Alibaba Cloud, and 2.8 million AI developers are active in the magic community in Alibaba Cloud. In the future, Alibaba Cloud will join hands with thousands of industries to promote AI innovation and share technology dividends."
AI moves from technology to industry
Tencent, Ali, Baidu and other Internet giants have poured into the big model market and launched their own big model products to continuously improve their technical level.
According to the Research Report of China Artificial Intelligence Large Model Map released by China Institute of Science and Technology Information, as of May 28th, at least 79 basic large models with a scale of more than 1 billion have been published in China. The number of large models developed in China has ranked second in the world.
At the recently held Yunqi Conference in 2023, Alibaba Cloud’s large model Tongyi Qianwen also ushered in the upgrade of 2.0, and became the third large model to surpass GPT3.5 after ERNIE Bot and Iflytek.
Li Yanhong, founder, chairman and CEO of Baidu, said at the Wuzhen Summit that the comprehensive level of Wenxin Big Model 4.0 is not inferior to Chat-GPT4.
Source: Tencent.com
Large-scale model technology is constantly improving, but the market is most concerned about whether the application of large-scale model can land and help the industry obtain stable commercial benefits.
At present, most of the models on the market are general models, which have the functions of natural language understanding and language generation, and are generally used in chat or entertainment scenes. But these scenes are not suitable for commercialization.
Compared with the general big model, enterprises prefer the vertical big model and focus on subdividing industrial scenarios, which is convenient for solving practical problems. In response to this actual demand, some companies with solid large-scale model technology have begun to explore the application of vertical scenes.
Xu Ran, CEO of JD.COM Group, introduced that JD.COM Big Model has been applied to many supply chains, such as business operation, customer service and after-sales, medical consultation and so on. She also pointed out that "the big model really realizes its own value, and it must be in industrial application."
Previously, Alibaba Cloud released a new model of Tongyi Qianwen 2.0, and also released a series of industry models to help developers use the power of artificial intelligence and algorithms in all walks of life to solve practical problems in combination with their own business scenarios.
Zhou Jingren, CTO of Alibaba Cloud, said, "Different scenarios have different needs. It is impossible for a model to serve all needs, and there is not only one corresponding way. Promoting the ecological prosperity of AI in China is the primary goal of Alibaba Cloud. "
A more open and prosperous AI ecosystem will help promote the commercialization of large models and bring them more sustainable commercial returns, thus achieving self-hematopoiesis.
In other words, by building a vibrant and innovative AI ecosystem, the big model will be able to better integrate into various industries and fields, exert its potential, create more value for enterprises and promote sustainable development.
Author | Pandora