After "lowering prices", Volcano Engine has released its next big move - establishing a retail industry model alliance.
On August 21st, the "AI Innovation Tour" series of activities organized by Volcano Engine landed in Shanghai after Chengdu and Xiamen, with a sea of people on site.
At the event, Tan Dai, president of Volcano Engine, announced that Volcano Engine would establish a retail model ecological alliance with 20 enterprises, including United Wumart Group, Yum! Brands (KFC, Pizzahut's parent company), McDonald's, Feihe, Haidilao, and Unimaginable Home. This is the third alliance launched by Volcano Engine after the Automotive Big Model Ecological Alliance and the Intelligent Terminal Big Model Alliance, and it is also an industry alliance that is closer to C-end users.
Tan Dai, President of Volcano Engine, stated that the concept of Volcano Engine is to collaborate with important companies and partners in industries where AI brings significant changes. This is the original intention of the alliance, and in the future, priority will also be given to industries with C-end scenarios.
In addition to building an industry wide model alliance, the Doubao big model is also experiencing acceleration.
At the Volcano Engine Powertrain Conference three months ago, Tan Dai announced the official debut of the Dou Bao Big Model family and announced that the enterprise level pricing of the Dou Bao Pro 32k main model has been reduced by about 99% compared to the industry, helping companies transform into AI with lower costs and easier to implement solutions, leading a wave of "AI price reduction trend".
Subsequently, the application of the bean bun big model continued to accelerate and upgrade. As of July, data shows that the daily use of tokens for the Doubao Big Model has exceeded 500 billion, and multiple Doubao vertical models, Volcano Ark, Button Professional Edition, HiAgent platform, and other products and cloud infrastructure have also received the latest upgrades.
The new trend of AI has come to this day. Chen Yuzhou, partner of Rolling AI, a consulting firm for enterprise intelligence innovation and AI scenario incubation, and many industry insiders have stated that AI big models provide more underlying AI technology capabilities, but do not directly create value. Only when they are truly applied to various industries and implemented in practical scenarios can they generate benefits. The most important thing in the future is to explore the actual demand scenarios of users.
As mentioned above, the volcano engine is exploring more practical application scenarios.
1、 Volcano Engine Rapid Iteration: Volume DAU Should also Volume Diversified Applications
At the WAIC (World Artificial Intelligence Conference) in July this year, a large number of top domestic and foreign AI technology companies, industry experts and scholars were gathered.
At the meeting, some people believed that in the era of AI, super capable applications are probably more important than super applications that only rely on DAUs. In the future, in fields such as healthcare, finance, education, manufacturing, transportation, agriculture, etc., millions of intelligent agents will be born based on the experience and demand scenarios of enterprises themselves.
The Volcano Engine is the player who not only "rolls" DAU but also rolls applications.
According to the data, the Doubao Big Model family has developed AI dialogue assistance packages, AI application development platforms, AI interactive entertainment cat boxes, as well as AI products such as Xinghui and Hema Aixue, based on the AI big model's abilities in text generation, voice interaction, sound replication, semantic understanding, massive data processing, and vectorized models.
These products have different functions and can meet the needs of people in different industries. From the perspective of the C-end experience, Doubao can currently communicate with Doubao users through voice interaction for multi scenario applications. For example, having Dou Bao act as an English speaking coach for interactive activities such as memorizing spelling, correcting pronunciation, chatting in real-life scenarios, and practicing English job interviews. Open the interactive entertainment app Cat Box, and after entering the character image, storyline, and other settings according to the prompt box, a colorful comic story can be generated within seconds to minutes.
At the event, a attending audience member was fixedly observing others using the AI application development platform in the booth area. The audience's name is Yang Zhiyuan, an AI product manager intern at Shuda Xian, a company dedicated to AI+education. As a sophomore, he is already a personal paying user of Volcano Engine.
Yang Zhiyuan started using Volcano Engine after its button platform was released in May this year. He introduced, "Stability is very important in the process of generating code in natural language. If we want to use it in a production environment, various factors such as concurrency and data security need to be considered. At that time, I designed a partial architecture for the product I was responsible for, which required the use of services from the Volcano Engine, such as AI inference access points, cloud storage, and concurrency processing. We had to build a part of our company's products on the Volcano Engine. After registration, the platform gave away 500 million tokens and cloud storage capacity as a gift, and we only became paying users after a good experience
The work of coding is still difficult for machines to completely replace. However, the significance of buttons lies more in replacing complex frameworks, using visual interfaces for agile development, which can to some extent replace existing large model application development frameworks such as LangChain, and greatly improve research and development efficiency, "said Yang Zhiyuan.
In real-life scenarios, the Dou Bao big model showed the highest performance improvement of 38.3% in six dimensions including role-playing. It has better performance in interactive expression in conversation scenarios such as chat companionship, audiobook synthesis, audio and video dubbing, digital person broadcasting, and voice customer service. At the same time, it can support recognition of multiple Chinese dialects, including Shanghainese, Minnanese, Cantonese, etc.
According to Tan Dai's on-site introduction, compared to its first release three months ago, the latest version of the Doubao big language model has improved its comprehensive ability by 20.3%, ranking first in China in multiple evaluations by authoritative institutions such as Zhiyuan and Shanghai AI Lab.
According to QuestMobile's report, the Doubao APP based on the Doubao big model is far ahead among similar products, with a monthly active user count of 27.52 million in the first half of the year, 2.43 times that of the second place.
In addition, the button platform also provides the function of agent development. According to the data disclosed by Zhu Jun, vice president of ByteDance products and strategy in May this year, more than 8 million agents have been created by users on Doubao.
Gao Xulei, Director of the Digital Finance Development Office of China Merchants Bank, once stated that even if big models are smart, they cannot directly affect the physical world. The emergence of intelligent agents has enabled large models to grow 'hands and feet', making them the best adaptation tool for AI to connect the 'last mile' of the physical world.
Chen Yuzhou previously told the media that accompanying companies to find AI scenarios for implementation is a long-term process, which requires testing and verifying feasibility in small-scale scenarios, training large models, and then debugging and iterating according to actual situations before it can truly be implemented. This process not only relies on AI big model technology, but also needs to help enterprises find the integration point with actual business.
2、 Value of the Big Model Alliance: Assisting AI Transformation in Thousands of Industries and Businesses
In the process of helping enterprises transform into AI and exploring practical business AI integration points, the solution of Volcano Engine is to establish an industry big model alliance.
Tan Dai stated on site that the implementation of large-scale model applications not only requires technological maturity, but also suitable commercial organizations and ecosystems. To this end, Volcano Engine has established a retail big model ecological alliance, covering many fields such as supermarkets, catering, food, clothing, etc., aiming to redefine the purchasing, experience, product, supply chain and other links with AI interaction as the core, enhance the turnover efficiency of the retail industry, and improve consumer shopping experience.
Tan Dai believes that the reason for choosing to establish an alliance in the retail industry again after opening the two major alliances of automobiles and intelligent terminals is that "automobiles and mobile phones are both terminals, and the interaction of AI on terminals is very natural and appropriate, so let's start. This time, we chose retail because we feel that the retail industry covers enough people and data, and we believe that it can bring about great changes and is worth innovating. We hope that every alliance has meaning
According to the data, the automotive big model ecosystem alliance previously established by Volcano Engine has also welcomed the second batch of members, including manufacturers such as Lynk&Co, Geely Galaxy, Geometry Motors, and SAIC MG. In addition, the Dou Bao Big Model has also made new progress in industries such as education and gaming.
Shen Wenjie, the head of the Marketing Digital Center of Geely Automobile Group, introduced that with the help of the Big Bean Bag model, Lynk&Co has achieved intelligent invitation of agents, AI training, and more; Multiple model-based functions such as internal training, AI sales assistant, intelligent management report, and AI car manual comprehensively assist in improving sales capabilities.
Regarding the direction of the next alliance, Tan Dai stated that all economic consumption ultimately relies on the C-end, and the ultimate buyers are the C-end. B2B will also end up in the C-end. At the same time, the C-end scenario has a high degree of digitization and a lot of data, making it easier to implement in both the intelligent cloud and AI industries. The Volcano Engine will definitely prioritize industries with C-end scenarios as a whole.
Tan Dai stated that the technological transformation of large models will have a profound impact on the intelligence process of various industries. Volcano Engine is committed to providing customers with "stronger models, lower prices, and easier to implement" full stack AI solutions, injecting new momentum into the future development of the industry.
But whether using AI big models or intelligent agent applications, whether for individual or enterprise users, embracing AI means embracing the 'future'. Dr. Zhang Wenzhong, founder of Duodian DMALL and Wumart Group, pointed out that "we must fully embrace AI, not only for a better future, but also for survival".