Home >Technology peripherals >AI >Four major areas for AI evolution! What exactly was announced on the UOS Technology Open Day?
UOS AI, browser AI assistant, mailbox AI assistant, natural language global search, and ChangWu Online... On October 24, a number of the latest AI technologies, applications, and products were unveiled together with large models. 2023 Tongxin UOS Technology Open Day and deepin Meetup Beijing Station.
Zhang Lei, senior deputy general manager, CTO and chairman of Deepin Community (deepin) of Tongxin Software, delivered a speech on the theme of "AI2.0: AI sinks, OS moves forward". In addition, from the terminal operating system production line of Tongxin Software Programmer representatives from , Changsha Online, server operating system and cloud computing production lines appeared to share.
"The explosion of generative artificial intelligence has opened up unlimited innovation space for the operating system in the second half."
As the cornerstone of the digital age, the operating system provides a running and supporting platform for all computer software. As a kind of software, large models themselves also run on the operating system.
In Zhang Lei’s view, the access to large models brings more possibilities to the operating system. The operating system can better understand user behavior and improve user experience:
First, the general large model explores more application scenarios through the operating system. The general large model has powerful language understanding and processing capabilities and can provide users with an intelligent experience;
Secondly, the emergence of large models has brought about a new interaction model. In the future, each user will have an exclusive AI assistant, which will run on the operating system;
Finally, in the future, AI will be one of the basic capabilities of the operating system. Partners can directly call the AI capabilities provided by the operating system to release the value of the platform.
In September, the Deepin Community operated by Tongxin Software officially announced that deepin has become the first open source operating system to be connected to a large model, and released UOS AI as scheduled, liberating the productivity of thousands of industries and unleashing everyone's imagination. .
"The advent of UOS AI fully reflects the system and efficiency advantages of Unongxin UOS." Zhang Lei believes that accessing large models is only a decisive step towards the second half of the operating system, and the ultimate goal is to deeply integrate and implement it with actual scenarios. converted into productivity.
He said that Tongxin UOS is not only enabling AI from the aspects of underlying XPU driver, runtime optimization, AI framework support, etc., but also working with many large model partners to integrate large models into the operating system. Next, Tongxin Software will also explore multiple technical points such as large models and AI native applications, natural language interaction compatibility, and data security to create the next generation operating system and innovation ecosystem.
"Without rich applications built on top of large models, large models will be difficult to implement."
At the scene, Zhang Peng, programmer representative of the terminal operating system production line of Tongxin Software, shared the theme of "Application first, three self-developed applications are connected to UOS AI". He pointed out that the UOS UOS desktop version links tens of millions of individual users and nearly 10,000 ecological partners, and is an important battlefield for UOS AI to reshape the application experience.
The currently released UOS AI can provide cloud large model docking services and local model docking services for upper-layer applications. At the same time, it provides centralized management capabilities for large models and multiple scalability capabilities based on prompt words, large model configuration, and source code level.
In terms of cloud large model docking, UOS AI has completed the docking of Baidu Qianfan, iFlytek Spark, Zhipu ChatGLM, 360 Intelligent Brain and other large models. The upper-layer application can complete multiple tasks through a unified interface. Access to each model eliminates the need for application software to adapt to each model. Currently, corresponding services are provided for the browser and mailbox of UOS UOS.
In terms of local models, it currently supports natural language understanding models and OCR models for upper-layer applications. With the help of these local models, upper-layer applications can quickly implement OCR recognition, natural language semantic understanding and splitting. Currently, it has been used for Tongxin UOS The global search application provides support and implements functions such as natural language search, image text search, etc.
“When large models become a new generation of “infrastructure”, smart documents are the first to usher in change, and are also the fastest to launch product categories for the market.”
Kuang Liang, programmer representative of Tongxin Software Changdu production line, said when sharing that document types can be divided into flow documents (suffix files such as doc, docx, wps, etc.) and layout documents (suffix files such as pdf and ofd). , streaming documents have the characteristics of free typesetting and can use large natural language models to complete content generation. Therefore, office software has chosen to take the lead in testing the waters in its online document products.
As the world’s first third-party application to access UOS public AI underlying services, Changsha Online has realized three content generation capabilities:
First: Outline of the document from 0 to 1
Everything is difficult at the beginning. If you want to improve office efficiency, the key is to take the first step. Create a new document in Changsha Online, call UOS AI, select the document outline, enter the title, and UOS AI can generate a complete article outline.
Second: Document Intelligent Assisted Creation
With the help of UOS AI, AI continuation, translation, language error correction, text polishing, content expansion and summary, etc. can be realized during the editing process. Until you get satisfactory results, click Finish to insert it into the text of the article with one click.
Third: AIGC multiple rounds of dialogue
Built-in real-time dialogue AI tool in the document, which can conduct multiple rounds of dialogue. Users can ask questions at any time within the document, changing the interactive interface and method of traditional online documents. It is equivalent to having an intelligent document assistant with real-time assistance, making creation more focused.
"We hope that before users start any creation, there will be a brainstorming session with AI to inspire creative inspiration and get into work faster." Kuang Liang said that when going deep into industry fields and professional scenarios, UOS AI will also It can provide users with more goal-oriented assistance.
“Facing the challenges of the artificial intelligence era and effectively initiating corporate transformation requires not only decisive courage, but also tailor-made change strategies.”
At this technology open day, Wu Yanping, a programmer representative of the Unison software server operating system and cloud computing production line, emphasized that in order to effectively promote the comprehensive application and in-depth development of artificial intelligence technology, the Unison UOS server version has continued to innovate R&D and technological breakthroughs have successfully solved a series of major challenges related to software and hardware compatibility.
In terms of software, Tongxin UOS server version has been adapted to many mainstream artificial intelligence frameworks, such as TensorFlow, PyTorch, MindSpore and PaddlePaddle, etc., which not only covers artificial intelligence development needs, but also provides a rich and diverse tool set to enable development Readers can conduct in-depth research and exploration in natural language processing and computer vision.
In terms of hardware, Tongxin UOS server version also provides support for a variety of artificial intelligence devices, such as NVIDIA GPU, Atlas training card and inference card, etc., which not only has excellent performance, but also has powerful computing power to meet the needs of Diverse artificial intelligence computing needs in various application scenarios.
In terms of cloud native, the Unitrust UOS cloud native version (abbreviation: "Unongxin Youyan") and the Unitrust Container Cloud Management Platform (abbreviation: "Unongxin Youque") are undergoing comprehensive upgrades simultaneously, integrating the mainstream in the industry New concepts and technologies such as DevOps, microservices, service grid, CI/CD, etc. accompany customers every step of the way to cloud native.
From being the first to launch UOS AI to integrating into thousands of industries and entering thousands of households, the challenges and explorations in the middle are often the most overlooked part. As the "Leader of Chinese Operating Systems", Tongxin Software hopes to stick to the long march of technology and create a foundation for innovation, making Tongxin UOS ubiquitous, making AI omnipresent, and truly giving the world a better choice!
The above is the detailed content of Four major areas for AI evolution! What exactly was announced on the UOS Technology Open Day?. For more information, please follow other related articles on the PHP Chinese website!