Shulou(Shulou.com)12/24 Report--

With the revolutionary breakthrough of AI large model capability, the development of large model technology has also ushered in a peak. In the next few years, AI technology will continue to focus on "improving large model capability" as an important development direction, enabling key technologies to improve, and driving large model application scenarios to land in various industries.

However, we should also see that behind the blooming of a hundred flowers in the domestic large model industry, there are still many problems: the opening of large models and the deep application of AIGC put forward higher requirements for data, algorithms and computing power, but the scarcity of AI computing power is the biggest headache for everyone; at the same time, the development of large models also poses more challenges to security and privacy; more discussion is how large models can be better commercial innovation?

Against this background, on December 7, Huawei Yunlian's big model sharing meeting with the theme of "unlimited intelligent emergence and innovation" opened in Beijing. This event gathered nearly half of the domestic large model head enterprises, and made an in-depth interpretation of the industrial trends, business practices, laws and regulations of the development of the big model, as well as the variables and possibilities of the new technology of the big model. Participants had an in-depth discussion with "how to innovate domestic large model scenarios and business models".

Group photo of the distinguished guests

IEEE / CAAI Fellow Tian Qi, chief scientist of Huawei Cloud artificial Intelligence, academician of the International Eurasian Academy of Sciences, delivered a speech at the event: the large model is the product of the times, which not only injects new impetus into the development of various industries, but also faces many challenges such as computing power, data, security compliance and so on. Huawei Cloud, as one of the earliest cloud service providers to lay out large models in China, firmly believes that large model technology will be widely used in more fields and bring more value and contribution to the development of human society. He also hopes to work with all the attendees to have in-depth exchanges through the platform of Lianchuangying, promote the innovation and application of large model technology, and contribute to the construction of an intelligent future world.

IEEE / CAAI Fellow Tian Qi, chief scientist of Huawei Cloud artificial Intelligence, academician of the International Eurasian Academy of Sciences

Five wonderful sharing in-depth Analysis of how the Big Model reshapes Enterprise Productivity and Innovation

In the process of digital intelligence transformation and upgrading of enterprises, the large model has become the key to promote enterprise innovation and productivity.

So what is the importance of large models in promoting enterprise innovation and digital transformation? Wang Feng, chief strategy officer of HuaWeiYun, points out that companies that can be reshaped by big models are the winners in the future. At the same time, with the exponential growth of enterprises' demand for computing power, infrastructure calls for all-round architectural innovation: chips need to continuously increase computing density to alleviate process bottlenecks; computing architecture will move from master architecture to peer-to-peer architecture. efficiency and energy consumption will be greatly improved; in terms of network switching architecture, orthogonal architecture will break through the engineering limits of high-density and high-speed interconnection. In the aspect of computing and storage collaboration, it is inevitable to build peer-to-peer high-performance cluster storage with global sharing.

Wang Feng, chief strategy officer of HuaWeiYun

Huawei's cloud model strategy is an important strategy of Huawei in the field of artificial intelligence, which aims to empower a wide range of industries with the ability of large models. From the point of view of the marginal cost and application of the large model era, Luo Hualin, vice president of Huawei cloud computing AI artificial intelligence field, after the artificial intelligence entered the large model era, although the marginal cost is getting lower and lower from the application point of view, the basic ability training at the bottom still needs a lot of computing power. He believes that the future model will be divided into two, one is to do the industry basic model, and the other is to focus on the application. Through multi-level and all-round cooperation with model partners, industry ISV and application enterprises, Huawei Cloud realizes and accelerates the prosperity and ecology of large models in the industry, and quickly falls to the ground to create value.

Luo Hualin, vice president of artificial intelligence for Huawei cloud computing AI

With the continuous development and application of the large model, the large model is also faced with more supervision and management to ensure its legal, safe and compliant operation. From the aspects of qualification access, intellectual property protection, network security and data protection, content compliance, algorithm compliance, marketing compliance, product compliance and so on, the transaction legal experts of Huawei Cloud Law Department introduce in detail the compliance issues and solutions that large model enterprises need to pay attention to.

Multimodal large model has become an important development direction of general artificial intelligence. Bao Fan, co-founder and CTO of Sheng Mathematical Science and Technology, introduced "the course of training the large model of Shengsheng Mathematical Science and Technology". He believes that the multimodal large model is developing along two directions: the expansion based on the large language model and the expansion based on the diffusion model, and shows the multimodal architecture based on the diffusion model developed by ourselves in the field, which can realize a variety of tasks, such as text, graphic, 3D, 3D and so on. At the same time, the quality and controllability of generation are guaranteed.

Bao Fan, co-founder and CTO of Sheng Mathematical Technology

The big model will reshape marketing productivity and accelerate the landing of innovation and value scene in the field of marketing. Martin, chief architect of Shenzhen Dongxinyun Technology Co., Ltd., believes that the application of large models in marketing will significantly expand the scale of creativity and content, improve the efficiency of content supply chain, accurately reach target users, and improve the efficiency of traffic conversion and realization. Dongxin Camp Competition Insight Model enables enterprises to achieve intelligent growth by providing real-time, credible and professional marketing content generation services.

Martin, Chief architect of Shenzhen Dongxin Cloud Technology Co., Ltd.

Two exchanges of High Combustion to explore the Development Direction of domestic large Model Innovation

This sharing session also opened a high-burning conversation between two big model celebrities.

In the round-table conversation on "variables and possibilities of new technologies for large models" hosted by Huawei Cloud artificial Intelligence Chief Scientist, academician of the International Eurasian Academy of Sciences and IEEE / CAAI Fellow Tian Qi, Zhou Ming, founder and CEO of Lanzhou Technology, Ru Liyun, co-founder and co-president of Baichuan Intelligence, and Yang Zhilin, founder of the dark side of the month. Several guests, including Cheetah Mobile Chairman and CEO and Orion Star Chairman Fu Sheng, discussed the development status of the large model, the opportunities and challenges to startups, as well as the threat of AI.

When talking about the development of large models from a global perspective, guests agreed that China's large model technology still lags behind OpenAI and Google in the United States, but in the application layer, some are already in the forefront, especially in the field of vertical large models in the industry. Everyone agreed that the current scarcity challenges faced by domestic large model companies in AI computing need companies such as Huawei to provide more computing support.

With regard to the development of China's large model industry, several bigwigs agree that although the industrial development will encounter various challenges, they are all full of confidence, and their unanimous view is that it is necessary to unite the entire industrial ecology and win-win cooperation. To solve the problems of "computing layer, data layer, application layer", China's large model industry will have great opportunities.

As for the important topics in the development of large models, such as data transaction and evaluation in artificial intelligence. Participants were of the view that in areas related to data security and privacy protection, there was a need to strengthen management and regulation, and to improve science and expertise in evaluation, so as to make customized assessments based on the actual needs of different areas.

Under the premise of such fierce competition in domestic large models, how should domestic large model manufacturers effectively achieve innovation? In the second exchange discussion on "domestic Big Model scene and Business Model Innovation" hosted by Hu Weiqi, Vice President of Huawei Cloud China, Tang Yi, chairman and CEO of impression Note, Wang Nan, co-founder and CTO of Jina A, Wei Qizan, technology partner of Xiaoku, Qi Ruifeng, co-founder of 010, and Guo Rentong, partner and head of product of Zilliz, discussed the topic of "big model innovation".

Guests believe that with the continuous development and popularization of technology, large models will become one of the basic settings in the future AI field, which will also bring more application scenarios and business opportunities to various industries. At the same time, the current large model also needs more technical breakthroughs and optimization to improve training efficiency, reduce costs, increase interpretability and so on. In terms of application, the large model needs more data and computing support and application scenarios to promote development. In addition, the application of the large model also needs to be customized and optimized according to the needs of specific industries and fields, in order to better meet the actual needs.

With regard to the much-focused topic of how startups in the field of artificial intelligence cooperate with large companies, participants believed that startups can share technology and resources to achieve win-win results when establishing partnerships with large companies. Startups can choose to cooperate with large companies to deeply segment the market and complete the last kilometer of the landing of large model applications. You can also choose to fill the large model niche and work with large factories to build a large model ecology, such as vector models, vector databases, model evaluation tools and so on.

Guests also had different views on the open source and closed source of the large model. Some participants were of the view that the open source model would have more room for development in the future, which could reduce costs and improve efficiency, while others considered that the closed source model was still the leading choice at this stage because it was more stable and reliable. and can provide better customer service and technical support. But what is certain is that at this stage, as the large model is still in the stage of rapid development, both open source and closed source have to accept the development of large model technology with an open mindset. It will bring subversive changes to our various industries and industries.

This large model sharing session of Huawei Cloud Union Camp provides participants with a platform for in-depth discussion of large model technology, through the sharing and exchange of technical experts, entrepreneurs and industry leaders, so that participants have a more in-depth understanding of the development trend, innovation scenarios and current challenges of the large model. In the future, Huawei Yun will continue to provide technical, computing and ecological support for the development of large models, and provide a solid "black land" for the development of China's large model industries.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.