Bytedance Beanbao large model released, Volcano Engine full-stack AI service helps enterprises intelligently transform

Volcano Engine President Tan Dai

If enterprises want to implement large models well, they face three key issues: model effect, inference cost, and implementation difficulty. Challenge: It is necessary to have good basic large models as support to solve complex problems, low-cost inference services to allow large models to be widely used, and more tools, platforms and applications to help enterprises implement scenarios.

——Tan Dai President of Volcano Engine

01.

Big bean bag Model debut

Massive use to polish the model

Model effect is the most critical challenge for the implementation of AI. Tan Dai pointed out that only through extensive use can a good model be polished. Currently, the Doubao large model processes 120 billion tokens of text and generates 30 million images per day.

In order to help enterprises implement large-scale model scenarios, ByteDance’s independently developed beanbao large-scale model will be officially provided to the outside world through the Volcano Engine. Enterprises can flexibly choose and implement quickly according to their own business scenario needs:

• 豆包universal model pro: ByteDance self-developed LLM model professional version, supports 128k long text, the entire series can be fine-tuned, and has stronger understanding, Comprehensive capabilities such as generation and logic are suitable for rich scenarios such as question and answer, summary, creation, and classification;
• Doubao general model lite: ByteDance self-developed LLM model The lightweight version provides lower token cost and lower latency than the professional version, providing enterprises with flexible and economical model choices;
• 豆包·Role-playing model: Personalized character creation capabilities, stronger context awareness and plot promotion capabilities to meet flexible role-playing needs;
• 豆bag·Speech synthesis model: Provides natural Vivid speech synthesis ability, good at expressing a variety of emotions and performing various scenes;
• 豆bag·Voice reproduction model: Sound 1 can be realized in 5 seconds: 1 clone, which highly restores timbre similarity and naturalness, and supports cross-language migration of sounds;
• 豆bag·Speech recognition model:Higher accuracy And sensitivity, lower speech recognition delay, supports correct recognition of multiple languages;
• 豆包·文生图model:More accurate text understanding ability, pictures and texts The matching is more accurate, the picture effect is more beautiful, and it is good at creating Chinese cultural elements;
• 豆包·Function call model: Provides more accurate function identification and parameter extraction Capability, suitable for scenarios where complex tools are called;
• 豆包·Vectorization model: Focuses on the usage scenarios of vector retrieval, providing core understanding capabilities for the LLM knowledge base, supporting multi-language.

Tan Dai introduced that as early as last year, the Doubao model (original name: Skylark) was launched within ByteDance, and it was also the first batch of large models to be safely registered through the large model service. one. More than 50 businesses within ByteDance have extensively used the Doubao model for AI innovation, including products with hundreds of millions of DAU such as Douyin and Toutiao. The Doubao APP, which is based on the large model of Dou Bao with the same name, ranks first among AIGC applications* in terms of downloads in the Apple APP Store and major Android application markets.

*Data source: QuestMobile

Zhu Jun, Vice President of ByteDance Product and Strategy

Zhu Jun, Vice President of Product and Strategy of Bytedance, introduced Bytedance’s product concept of building AI native applications based on the beanbao large model. He said that technology is evolving rapidly, but the core needs of users have not changed, such as the need to obtain information efficiently, the need to improve work efficiency, the need for self-expression, the need for social entertainment, etc. ByteDance has done a lot of study and exploration, hoping to find a suitable large-model application form to serve users in a natural interactive form and enter more people's lives and work.

Doubao APP is a large-scale model application that ByteDance has invested heavily in. According to Zhu Jun, the monthly active users of the Doubao platform have reached 26 million, and more than 8 million agents have been created.

While Doubao Big Model serves Byte, it is also co-creating with many enterprise customers of Volcano Engine. Tan Dai believes that enterprises need models with different characteristics in different business scenarios to achieve the best results and cost performance. To this end, Doubao Big Model provides a family of multi-modal models that enterprises can flexibly choose according to their own business scenario needs to achieve a positive cycle of model use and cost control.

02.

Everyone can afford it

A good model is

Tan Dai believes that reducing costs is a key factor in pushing large models forward to the "value creation stage". The price of Doubao’s main model in the enterprise market is only 0.0008 yuan/thousand tokens. It can process more than 1,500 Chinese characters at 0.8 cents, which is 99.3% cheaper than the industry.

Taking Doubao universal model pro-32k version as an example, the model inference input price is only 0.0008 yuan/thousand tokens. The pricing of models with the same specifications on the market is generally 0.12 yuan/thousand tokens, which is 150 times the price of the bean bag model. The price of large models has entered the "centi era".

Tan Dai said that the "roller" price of large models will help companies accelerate business innovation at lower costs.

03.

Ark 2.0 upgrade

Promote large model value creation

Last year, Volcano Engine released Volcano Ark, a one-stop large model service platform, hoping to help enterprises apply various models in an efficient and safe environment through the concept of model as a service. This year, Volcano Engine has comprehensively upgraded the Ark platform. The upgraded Volcano Ark 2.0 will significantly improve model effects, core plug-ins, system performance and platform experience, helping companies promote the value creation of large models.

In terms of plug-ins and tool chains, Ark 2.0 has upgraded the Internet plug-in, which provides the same search capabilities for Toutiao and Tik Tok, connects to massive high-quality data in real time, and uses text, images, voice, etc. Modal interaction method, and through the industry's leading intent recognition capabilities, greatly improve the model's information acquisition capabilities; upgrade content plug-in, provide Toutiao Douyin with massive content from the same origin, support multi-modal interaction, and provide intent-based Vertical content information retrieval, content timeliness retrieval is stronger, helping large models to deeply understand, retrieve and generate content; upgrade knowledge base plug-in to provide millisecond-level tens of billions of high-performance retrieval, second-level streaming The knowledge base index is updated and the bean bag vectorization model is embedded to improve the relevance and accuracy of search.

In addition to the above core plug-in upgrades, Volcano Ark 2.0 also comprehensively upgraded the system carrying capacity, security protection capabilities and algorithm service capabilities. In terms of system carrying capacity, it provides abundant GPU computing resources and super elasticity of kilo-card capacity expansion and contraction in minutes, ensuring business stability and cost controllability; in terms of security protection, it builds a trusted execution environment through a security sandbox. A multi-dimensional security architecture ensures data security; in addition, Volcano Engine provides professional algorithm team services to help customers unlock unique data value and enable large enterprise model applications to be implemented easily.

Pan Yuyang, Button Product Manager

In addition to good model effects, convenient plug-ins, and strong system performance, we also need to provide a simple and easy-to-use development platform. Let everyone become an AI application developer. Coze product manager Pan Yuyang introduced that Coze, as a new generation AI application development platform launched by ByteDance, has the advantages of low threshold, personalization, real-time, multi-modality, etc., and integrates massive AI resources and rich release channels. , one-click custom API service to help fill the last mile from large models to user scenarios.

Volcano Engine officially released the professional version of Button. Based on the ability of Kouzi to flexibly orchestrate intelligent agents, Kouzi Professional Edition further provides enterprise-level SLA and a variety of advanced features, making AI applications easier to implement and allowing enterprises to focus more on driving business growth through agent innovation.

Tan Dai said that Volcano Engine is committed to becoming the most open and leading cloud service platform in the AI ​​era. The newly upgraded Volcano Ark 2.0 will help implement large-scale model applications through stronger performance, better plug-ins, better services, and safe and trustworthy solutions.

04.

Extensive implementation

Fruitful results in multiple industries

Beanbao Big Model and Volcano Ark 2.0 have been implemented in many industries such as finance, automobiles, smart terminals, e-commerce retail, education and scientific research. Companies such as China Merchants Bank, Haidilao Hotpot, Super Orangutan, Ctrip Travel, Feichangzhun, and Liepin have also built smart bodies on buttons. Famous universities such as Fudan University and Zhejiang University have also built “AI teaching assistants” for courses and experiments.

Gao Xulei, Director of Digital Finance Development Office of China Merchants Bank

As a practitioner of AI transformation in the financial industry, China Merchants Bank is developing intelligent outbound calls, digital people, intelligent data assistants, etc. Explore and invest in hundreds of AI service scenarios. Gao Xulei, director of the Digital Finance Development Office of China Merchants Bank, said that the large model will bring transformation and upgrading to the bank's entire chain of business. Working together with Volcano Engine, the two parties will jointly build large-model application scenarios and high-performance infrastructure, transform existing systems and products, and generate a large number of AI-centric intelligent agent applications.

Li Chengjie, Vice President and Chief Digital Intelligence Officer of Mengniu Group

Li Chengjie, Vice President and Chief Digital Intelligence Officer of Mengniu Group, has evolved from the evolution of production tools and production methods , three aspects of the evolution of organizational methods introduce the digital intelligence process of Mengniu Group. Mengniu Group is cooperating with Volcano Engine to consolidate the technology base and build an intelligent matrix of AI nutrition experts, AI nanny, AI data analysts, product innovation assistants, hot spot creative assistants, etc. to enhance the front-end consumer experience and assist back-end operations. Personnel decisions.

Smart Terminal Large Model Alliance Release

At the meeting, Volcano Engine and OPPO, vivo, Honor, Xiaomi, Samsung, and ASUS announced the establishment of a smart terminal large model alliance. Applications such as OPPO Xiaobu Assistant, Honor MagicBook's YOYO Assistant, Xiaomi's "Xiao Ai Classmate", and ASUS laptop's Douding AI Assistant have all been connected to the large model service of the Volcano Engine.

Automobile Large Model Ecological Alliance released

At the same time, Volcano Engine joined forces with the China Electric Vehicles Association of 100 and Geely Automobile, Great Wall Motors, Jietu Motors, Saili More than 20 manufacturers, including Si and Zhiji Automobile, announced the establishment of the Automotive Large Model Ecological Alliance, which is committed to exploring new forms of human-vehicle interaction for a new generation and improving new AI experiences in all automotive scenarios.

The two major alliance companies OPPO and Jietu Motors also shared their own large model practices:

OPPO Director of Intelligent Assistant Department, Chief Structure of Xiaobu Assistant Shi Wan Yulong

OPPO Intelligent Assistant Department Director and Xiaobu Assistant Chief Architect Wan Yulong shared the large-scale model application practice of OPPO AI mobile phones. He said that Xiaobu Assistant has currently covered 350 million terminal devices. , the number of monthly active users exceeds 150 million. On Volcano Ark, OPPO built intelligent agents such as universal question and answer, smart girl, and interviewer through intent recognition models, network summary models, personalized character models, and beanbag models, bringing brand-new AI to end users. experience.

Liu Hang, head of digitalization of Jietu Automobile

Liu Hang, head of digitalization of Jietu Automobile, said that Jietu Automobile hopes to expand its service capabilities through large models , gain insight into the real needs of users, thereby improving sales efficiency and capabilities. Working together with Volcano Engine, the two parties co-created around customer service scenarios. The smart customer service AI Xiaojie created not only saves manpower, but also helps customer service personnel quickly obtain information and serve customers. In the future, Jietu Motors also hopes to further realize intelligent upgrades in cockpit, decision-making, and marketing through large models.

05.

Multiple AI applications released

Quickly implemented and ready to use out of the box

In addition to the large model service platform, AI applications are also a key component of the Volcano Engine full-stack AI service.

Last year, Volcano Engine proposed a new model for enterprise digital intelligence upgrade - data flywheel. This model takes data consumption as the core driving force, fully integrates enterprise data flow into business flow, and realizes the flywheel effect of data assets and business applications. In order to make enterprise data consumption more inclusive and convenient, Volcano Engine Intelligent Data Insight DataWind officially released the AI ​​assistant ChatBI. Users can generate indicators through natural language interaction for chart production, data query, and in-depth data analysis. ChatBI It supports IM interaction, can be applied on multiple terminals, and performs self-service analysis anytime and anywhere. This will significantly lower the data consumption threshold, improve reading and data usage efficiency, and provide power for accelerating the operation of the data flywheel.

Facing marketing scenarios, Volcano Engine has been upgraded to Intelligent Creation Cloud 2.0, becoming a one-stop enterprise content marketing growth solution from content creation, matrix distribution, advertising to data insights; not only provides multi-modal The ability to understand and generate helps increase video creation efficiency by 25 times, and the launch of Douyin topics, POIs, mini programs and other marketing tools has increased store page exposure by 600%. Currently, 3,000+ companies have started AIGC video marketing growth on Volcano Engine.

For sales scenarios, Volcano Engine has released Sales Copilot, a sales AI assistant that can efficiently follow up on customer needs and assist sales in answering complex product questions at any time. In addition, it can simulate different styles of customer drills through role-playing, and provide excellent speaking skills to improve the quality of sales communication.

Volcano Engine has been committed to building an open and shared ecological concept. Since the release of Wanyou Plan in 2021, it has helped 100+ SaaS companies achieve business growth and achieved nearly 10,000 customer signings. Doubao Enterprise Service Alliance is a special plan of Wanyou Plan for the implementation of large models. Wanyou Plan Doubao Enterprise Service Alliance will provide a package of support plans to help SaaS companies achieve AI upgrades and provide industry customers with more out-of-the-box AI applications.

06.

AI infrastructure is further upgraded

Reliable guarantee of cloud computing power

In the era of large models, cluster management complexity, network performance, and stability are key challenges to cloud infrastructure. In order to deal with the above challenges, Volcano Engine provides a set of industry-leading solutions around large cluster construction, management, performance optimization, and stability guarantee based on the resource scale advantages and internal practices of internal and external common pools.

NVIDIA Vice President of Solutions and Engineering Lai Junjie

At the meeting, the Volcano Engine officially released veOmniverse, which provides enterprises and developers with a new platform based on NVIDIA Omniverse. The cloud development platform helps enterprises and developers carry out efficient design and collaboration around workflow in the interconnected virtual world, and explore and create intelligent and realistic virtual worlds.

In addition to providing customers with high-stability and high-performance cloud infrastructure, in response to local deployment needs, Volcano Engine has also released a new hybrid cloud veStack intelligent computing version, with Wanka cluster networking, 3.2 T The ultra-large-scale advantage of high-performance lossless network; the ultimate performance supported by end-to-end engineering, able to achieve 97.78% training acceleration ratio and minute-level fault discovery and self-healing; it has multiple heterogeneous capabilities and can be adapted to more than ten types of GPUs , supports mainstream domestically produced GPUs.

In addition to the intensive computing performed by the GPU, a complete AI application also needs to continue to reduce costs and increase efficiency on the existing CPU load to ensure that business can be carried out easily and accelerate innovation. From the four levels of computing, storage, network, and operation and maintenance, Volcano Engine fully supports the stable operation of customers' cloud businesses.

End

From technological innovation to commercial implementation, large models have provided powerful momentum for the AI ​​transformation in various industries. The Volcano Engine is based on ByteDance's practical experience in the field of large models. It has polished good models through large-scale usage. It makes large models affordable for everyone with its ultimate cost-effectiveness driven by technology. It also accompanies customers through its rich implementation experience and algorithm services in various industries. Grow together. In the future, Volcano Engine's full-stack AI services will continue to help enterprises implement AI transformation and stimulate business growth potential.

The above is the detailed content of Bytedance Beanbao large model released, Volcano Engine full-stack AI service helps enterprises intelligently transform. For more information, please follow other related articles on the PHP Chinese website!