Leading the AI ​​revolution, Jiuzhang Yunji DataCanvas launches AIFS+DataPilot

On June 30, the "New AI · New Data · New Software" Jiuzhang Yunji DataCanvas new product launch conference was held in Beijing. More than 50 companies came from Tsinghua University, China Academy of Information and Communications Technology, IDC China, Xinhua News Agency, etc. Guests from universities, institutions and media witnessed the birth of two transformative AI product series, "AIFS Artificial Intelligence Basic Software" and "DataPilot Data Navigator" with Jiuzhang Yunji DataCanvas Company.

The technological breakthrough of large-scale AI models opens a new perspective, allowing people to immediately look forward to a colorful future. When belief becomes a force, it will promote immeasurable leaps in development. At the press conference, Fang Lei, chairman of Jiuzhang Yunji DataCanvas Company, explained the unique world view of large models from the perspective of AI technology companies.

“The era of large models requires complete infrastructure upgrades, rather than relying on a single large model to solve all problems; the implementation of large models will solve more difficult problems and cause more far-reaching impacts, and it is not easier than small models.”

Fang Lei pointed out that the complete infrastructure includes three major elements: computing power, data and basic software. Whether large models can achieve cross-era development depends on the frequency of progress of the three major elements.

According to current trends, the continuous improvement of computing power heralds the advancement of hardware, which means that the resource gap will no longer be a problem. There is still broad room for effective storage, calculation and circulation of data. In the real world, there are many independent data territories among industries, enterprises, and professions. The huge amount of data and the difficulty of connecting data territories indicate the emergence of general large models. Difficulty of landing. The implementation of large models will be earlier and more reflected in vertical large models such as industries and enterprises. Similarly, the number of vertical large models will greatly exceed general large models. As for basic software, there are large differences in performance and cost. The place where innovation vitality is most concentrated is the unified optimization space of software, models and hardware, Fang Lei pointed out.

AI technology in the era of large models still requires the integration of the “last mile”. The most active place for innovation is the huge space for unified optimization of software, models and hardware, Fang Lei pointed out. Basic software with strong flexibility, using an open and flexible white box model, and driven by professionals who are proficient in business will quickly achieve breakthroughs in the last mile. The AIFS & DataPilot product system launched at this product conference is a brand-new technological achievement with this goal.

AIFS, use the power of AI to explore the boundaries of model application capabilities

AIFS (AI Foundation Software) is Jiuzhang Yunji DataCanvas’s answer to comprehensively building AI capabilities in the New AI era dominated by large models. It is also the latest upgrade of the DataCanvas product family.

As an industry-leading artificial intelligence application construction infrastructure platform, AIFS covers the training, fine-tuning, compression, deployment, inference and monitoring of large models as well as the full life cycle process of small models. It provides data scientists, applications Program developers and business experts provide a set of tools that allow people in different roles to collaborate with each other to easily process and use data to develop, train and deploy models of any scale.

As an artificial intelligence basic software system, AIFS mainly includes DataCanvas Alaya Jiuzhangyuanshi large model, DataCanvas APS machine learning platform, DataCanvas BAP business-oriented automatic modeling platform, open source DAT automatic machine learning software, open source YLearn causal learning software, etc. A series of fully open, highly automated, and highly collaborative software tools provide one-stop support for users to independently build “big and small” models throughout the life cycle.

The newly released DataCanvas Alaya Nine Chapter Yuanshi large model has the characteristics of "general knowledge industry" series model matrix, multi-modal large model, optimized training mechanism and friendly open source protocol management. In terms of open source support, Jiuzhang Yuanshi not only supports the Apache 2.0 protocol, but also provides users with white-box models, which is really outstanding in the current large model industry. Yu Jiangang emphasized that this is the company's insistence on the "openness" of the product, aiming to give users greater freedom in AI innovation capabilities in order to accelerate the application of large models in diverse business scenarios.

Multimodality is the next important technical link in the large artificial intelligence model. Miao Xu, chief AI scientist of Jiuzhang Yunji DataCanvas Company, focused on the technical route of the multimodal direction of the Jiuzhang Yuanshi large model. In order to support various industries, Jiuzhang Yuanshi vigorously develops the integration and alignment between structured data and unstructured data, so that large models can not only use text information, but also use industry databases and a large number of sensors that are not easy to process. Sequence makes intelligence closer to business scenarios. Furthermore, Jiuzhang Yuanshi also provides fine-tuning technological innovations for the construction of enterprise-owned large models. It improves efficiency by dividing complex fine-tuning tasks into different sub-tasks, making the large model customization process easier and more comfortable. .

DataCanvas AIFS is not a small step from small models to the era of "big and small" models. It is a big step that represents the integration of core software capabilities and white-box models in the era of large models. AIFS can also achieve huge empowerment for the application side, build personalized and independent large models for future enterprises, and integrate the large models with the small models accumulated in the past and then apply them to the business, pressing the "confirm button".

DataPilot, a new era tool for swimming in the "vector sea" of data

AI and Data have always been closely related. In the past decade or so, data has often been regarded as the raw material and foundation of AI. We call this era the era of data as the architecture of AI. The AI ​​reverse empowerment of data is a sign of the new New Data era, all thanks to the emergence of large models.

When big breakthroughs in data and AI capabilities collide, how will its future change? DataPilot is giving answers to the world, and "vector sea" has become the key word of the answer.

DataPilot, a new data processing paradigm, is a new generation of data architecture tool product based on large models independently developed by Jiuzhang Yunji DataCanvas. Taking full advantage of the general text understanding and generation capabilities of DataCanvas Alaya's nine-part metadata model and its fine-tuning and optimization in the data field, DataPilot can help users achieve intelligence and automation in the entire life cycle of data modeling.

The "Vector Ocean" is the ultimate form of data development creatively proposed by Jiuzhang Yunji DataCanvas Company based on years of research and practice in the database field and combined with the development direction of vector data.

Zhou Xiaoling, vice president of Jiuzhang Yunji DataCanvas Company, introduced that DataPilot’s features include multi-modal “vector sea” data architecture, on-demand automated data integration, code generation, process orchestration and analytical calculation, as well as natural language-based data acquisition, Analytics and machine learning modeling capabilities. DataPilot can significantly reduce the technical threshold of data integration, governance, modeling, calculation, query, analysis, and machine learning modeling, reduce the cost of data-driven business development, and accelerate the process of digital innovation.

Based on the concept of "vector sea", DataPilot includes various data software such as the DataCanvas RT real-time decision center platform and the open source DingoDB multi-mode vector database, allowing users to have the real-time and multi-mode capabilities that are urgently needed when AI technology breaks through. Stateful data capabilities.

In the era of vector sea, DingoDB, as a powerful multi-modal vector database, will become the engine. It combines the features of data lakes and vector databases and can store data of various types (such as key-value, PDF, audio, video, etc.) and various sizes. Through DingoDB, users can build their own data "vector sea", whether it is structured or unstructured data, and can complete multi-modal data analysis and scientific calculations with only one set of SQL.

From Software to Thought-ware, a new form of software evolution

While empowering each other with Data, AI is also having a profound and even subversive impact on the evolution of software forms. The era of New Software has officially begun.

Empowered by new breakthroughs in AI technology, the world is experiencing a leap from the "Software" era to the "Thought-ware" era. Traditional "software" is a process of continuous iteration around the three links of demand analysis, product design, and code implementation. It is a paradigm of "passive response to demand". "Thinkware" is a new paradigm of software evolution with "thinking" as the core.

Yang Jian, Chief Architect of Jiuzhang Yunji DataCanvas Company, further explained "Thinkware" at the press conference. He said that "Thinkware" has the ability to think independently, control actions, and proactive self-evolution. ability.

The ability to think independently enables thinkingware to understand user intentions and combine it with action capabilities to provide users with beneficial solutions. The ability to think and act autonomously will raise a series of concerns about the legality, safety, and compliance of system capability boundaries, so such system action capabilities must be "controlled." In addition, thinkingware must also have the ability to actively self-evolve, which is an ability to actively learn and self-optimize. During operation, it can learn independently in the interaction with users and the environment to achieve error correction, demand adaptation, and A series of evolutionary processes such as capability precipitation and optimization.

Yang Jian revealed that Jiuzhang Yunji DataCanvas Company has begun to explore the new software "Thinkware" and demonstrated the "Thinkware" experimental product TableGPT at the press conference.

TableGPT follows the concept of "what you need is what you get". Users only need to describe problems and needs in natural language. There is no need to enter complex commands or perform manual algorithm selection. TableGPT can automatically understand user intentions and select appropriate statistical analysis, Machine learning methods are used to complete automatic modeling on private domain data, and then feedback the data analysis results required by the user and explain them, and can also give suggestions for subsequent data mining. Users do not need to worry about or be exposed to any specific technical details, which greatly reduces the difficulty of learning and using.

This kind of language-based interaction with artificial intelligence software allows anyone to easily gain insights from data. The minimalist, novel interactive experience, and unexpected interactive results will vigorously promote the democracy of data analysis. ization and popularization. TableGPT points out the direction of the integration and development of data science and artificial intelligence software, although it is still in its infancy, Yang Jian said.

From AIFS to AIFS

The artificial intelligence industry has shown a new round of explosive vitality under the upsurge of national participation in AIGC interaction. No matter what branches the AI ​​industry develops in, high-performance basic software and data architecture will always serve as the foundation. Play an important role in the basic core. The two major product systems released this time continue to implement Jiuzhang Yunji DataCanvas' core product concept of "open, automatic, and cloud-native" and will play an important role in the rapidly changing new world of AI.

With the mutual empowerment of AIFS, DataPilot, and Thought-ware, Jiuzhang Yunji DataCanvas’ software tools and solutions will continue to empower a wide range of industries, continue to integrate cutting-edge AI innovation technology, and help them in large models The era is accelerating the realization of independent digital intelligence upgrades and large-scale application of AI.

Facing the future, Jiuzhang Yunji DataCanvas Company will adhere to the strategic top-level planning of "cloud in cloud" (AI cloud among thousands of clouds), and cooperate with partners such as cloud manufacturers and intelligent computing centers to provide one-stop services and complete A gorgeous transformation from AIFS (AI Foundation Software) products to AIFS (AI Foundation Service) services.

The above is the detailed content of Leading the AI ​​revolution, Jiuzhang Yunji DataCanvas launches AIFS+DataPilot. For more information, please follow other related articles on the PHP Chinese website!