In "The Structure of Scientific Revolutions", Thomas Kuhn focused on the role of new paradigms in scientific revolutions and industrial revolutions. In a deep technological transformation, industrial development cannot be immersed in the old development model, but needs to actively practice paradigm transformation and complete a leap on the basis of new technologies and solutions.
This is especially true for industry intelligence. Today, the combination of AI and industry has become the consensus of all walks of life. However, in the traditional sense, a series of bottlenecks are emerging in the way of empowering the industry with general APIs or fully customized solutions. Whether new methods and new paradigms can be found for the intelligent development of the industry is becoming a concern of all walks of life.
On September 2, during the 2022 World Artificial Intelligence Conference (WAIC), the Shengteng Artificial Intelligence Ecological Conference co-sponsored by the New Generation Artificial Intelligence Industry Technology Innovation Strategic Alliance (AITISA), China Artificial Intelligence Industry Development Alliance (AIIA) and Huawei Held simultaneously online and offline. At the conference site, Huawei, together with partners such as SenseTime, CloudWalk, AsiaInfo, Extreme View, Yuntian Lifei, Zhongke Hongyun, and Bohan Intelligent, jointly released an industry scenario-based solution based on Ascend AICE. Energy cities, education, manufacturing, finance and other industries. In the AICE empowerment industry solution, we can find a new paradigm and a new starting point for the intelligent development of the industry.
The AI empowerment solution that is highly integrated into the industry mechanism, deeply meets the needs of the industry, and achieves the continuous evolution of industry wisdom is accelerating to become a reality.
In the application era, where is industry intelligence going?
Not long ago, the Ministry of Science and Technology issued the "Notice on Supporting the Construction of a New Generation of Artificial Intelligence Demonstration Application Scenarios", announcing the first batch of 10 scenarios of artificial intelligence demonstration applications, emphasizing the need to give full play to the role of artificial intelligence in empowering economic and social development, and form a group of Reproducible and scalable benchmark demonstration application scenarios. This policy releases a signal for the country to accelerate the innovation of artificial intelligence scenarios and industrial development. At the same time, it also puts forward new requirements for the application of AI in various industries and the deep activation of AI capabilities.
Overall, there are several options for industrial application of AI. One is to directly apply general AI capabilities, but this method is too simple to meet the widely existing differentiated intelligence needs in the industry; the other is a fully customized solution, but its development cost is too high and the development cycle is too long. It is difficult to be used by small and medium-sized enterprises and industries with weak digital foundations; the third is to empower the intelligent development of the industry based on the development platform. This method requires the enterprise itself to have certain development capabilities, but the AI talent and capacity reserves in the industry are often insufficient.
In the process of moving from the algorithm era to the application era, industry intelligence is ushering in new developments. Just as people need to grow again in social work after learning basic knowledge in school, AI also needs to penetrate into the production system of the industry, and continuously combine actual scene data for incremental training and iterative learning. Such an industry production system needs to meet the three requirements. Item Features:
1. Computing with the data, building industry computing power infrastructure around the data deployment environment, ensuring the safe acquisition of local data, so as to realize the combination of computing power and data.
2. Integrating training and pushing, building an integrated platform for training and reasoning, and building a complete production system to ensure that the cyclical production process from training to reasoning is no longer separated.
3. Cloud-edge collaboration enables difficult case data to be efficiently sent back to the center for training, and the iterative algorithm is quickly deployed to the inference device, truly generating value for industry scenarios.
Faced with such changes in demand, AICE based on Ascend AI brings a series of problem-solving ideas to support ecological partners in incubating more high-quality industry solutions, and deepening industry scenarios and industry wisdom.
Fresh start:
The differentiated value that AICE brings to industry intelligence
Artificial Intelligence Convergence and Empowerment Platform (AICE for short) is a training-reasoning integrated platform with autonomous evolution capability based on Ascend AI basic software and hardware. From the perspective of technical architecture, AICE is currently the industry's most comprehensive hardware and software infrastructure platform that can meet the needs of users in the entire process from training to inference deployment. AICE deeply integrates the technical capabilities and product capabilities of the Ascend system, and can provide the industry with central and edge AI computing power resources, as well as software such as the Euler operating system, the heterogeneous computing architecture CANN, and the AI framework MindSpore; the business enablement layer , together with AICE's platform partners to build an AI training platform and reasoning platform, carry out algorithm updates and incremental training, and build a training and push integrated platform with autonomous evolution capabilities; the algorithm application layer, through global perception, knowledge computing, robots and other engines to achieve global intelligent.
The comprehensive software and hardware infrastructure allows AICE to help various industries realize the full opening of traffic in the AI era, and become a new starting point for industry intelligence. On the whole, AICE can generate triple differentiated value:
Open platform and prosperous ecology: Based on the connection of platform interfaces, AICE can quickly adapt with partners to meet more complex industrial intelligent scenarios. At present, AICE has gathered the capabilities of more than 20 partners, providing 150+ algorithms with an accuracy of 90%+.
Integration of training and push, cloud-side collaboration: By deploying an integrated environment for training and reasoning, AICE can realize continuous optimization of the model. It supports the central node to send the model to the edge node for reasoning, and the edge site then sends the data back to the center to achieve agile iteration. Such training-deployment capabilities can maximize the satisfaction of industry AI application and development scenarios, and realize the deep integration of AI infrastructure and industry needs.
Full-stack independent innovation: From root technologies such as AI hardware, chip-enabled software, and AI framework, to upper-layer applications such as application enablement and development platforms, Huawei has carried out full-stack independent innovation. It can help enterprise users realize safe, credible, and reliable AI implementation, and promote the construction of a localized AI ecosystem.
Paradigm shift:
AI integrates into the industry, AI achieves the industry
In the emergence of AICE's empowering industry solutions, we can see that a new industry intelligence paradigm is being generated. Compared with traditional AI empowerment methods, AICE takes more into account the development characteristics of the industry itself, the needs of technology implementation, and the necessity of ecological openness and industrial collaboration. The core of this development paradigm lies in switching from the algorithm-centered empowerment solution to a new development paradigm centered on user needs and based on industry wisdom.
AICE empowers industry solutions, which can bring triple industry impacts to the extensive intelligentization of industries in the future:
1. Fully satisfy the industry's own intelligent development path, and give the initiative to develop AI to the industry and enterprises themselves.
2. Avoid the problems of repeated construction of intelligent systems and AI capabilities, and the waste of a large amount of resources, and achieve long-term sustainable industry AI capability building.
3. Through ecological collaboration and industrial linkage, effectively solve the complex AI needs emerging in the industry, and ensure that the core production units of the industry can also integrate AI technology.
In the "White Paper on Artificial Intelligence Convergence Empowerment Platform" released by Huawei and ICT Institute, we can see real cases in the intelligent development of the industry, and how all parties can define their own positions in AICE empowerment industry solutions and seek value paths .
For example, Polar View creates urban governance solutions based on the Polar Stack platform, Polar Star platform and Ascend AI basic software and hardware; Yuntian Lifei creates deep-sea multi-dimensional big data based on the Deep Intelligence AI algorithm service platform and Ascend AI basic software and hardware Converged solutions; based on AsiaInfo’s general artificial intelligence platform and Ascend AI’s basic hardware and software, AsiaInfo creates a 5G+AI smart park solution, empowering smart cities with Al...Many industry partners and Huawei have jointly built a smart foundation and shared Win the future of the industry.
Based on the combination of AICE and the industry, a new industry intelligence paradigm is taking shape, and the implementation of AI has entered a new value field and development stage.
