A new round of industrial transformation is booming in the world today. New technologies such as artificial intelligence and digitalization have become the core driving force of the transformation, and are having a profound impact on the world economy, social operation and human life. At present, China has issued a series of planning guidelines for building a network power, digital China, and digital economy development. New-generation information technologies such as 5G network, big data, block chain and AI are gradually promoting the transformation of production modes and factors, making operators pay more attention to diverse businesses, quality experience and intelligent services. Therefore, it is an inevitable responsibility for the operators to build comprehensive intelligent information service across connections and promote intelligent and digital transformation to fulfill their corporate missions and social responsibilities.
ZTE is actively responding to China's digital transformation strategy, speeding up the transformation and upgrade of digital platforms, and building a digital service capability base from top to bottom. ZTE has independently developed the AI universal platform and enabled various kinds of industries through AI to build an ubiquitous AI capability system that helps traditional industries and products to move towards digital economy and promotes high-quality development of enterprises.
ZTE's AI platform integrates expert knowledge and evolutionary algorithm that can implement automatic data labeling, model structure search and hyper-parameter optimization. Through the Adlik reasoning tool chain, it compresses and compiles key models, and optimizes them during the reasoning operation, thus realizing computing power optimization for all scenarios. The one-stop self-evolving AI platform enables end-to-end AI-empowered business flows, and its advantages lie in algorithm, computing power and data:
Empowering is the essence of AI. AI applications are entering the era where scenarios are dominant. The success of AI application mode requires in-depth combination of technologies, closed-loop data, and scenarios (industry knowledge). The effect of cognitive intelligence is sensitive and related to scenarios. To obtain cognitive results, it is necessary to extract implicit knowledge or based on background relevance knowledge. In the post-deep learning era, in addition to the computing power, algorithms and data, AI-empowered scenarios and industry experts (knowledge) are particularly critical.
With the evolution of telecom networks towards virtualization, virtualized integration services are also transforming towards digital and data-driven services. The introduction of AI in integration services becomes an inevitable trend for future product development. AI can perform deep learning based on a large amount of training data generated in the process of network functions virtualization (NFV) integration, continuously upgrade the algorithm, and reverse-apply it to the scenario to become an integration capability multiplier. After analyzing the core requirements of operators for hierarchical decoupling, intelligent integration, and agile delivery in NFV integration, ZTE has proposed the plan of building AI-empowered intelligence capabilities of NFV integration in combination with the digital transformation and development path of virtualized integration services. AI is introduced upon the digital integration platform, and based on NFV components, self-intelligent hierarchical integration capabilities of native intelligence, hierarchical autonomy, and cross-layer collaboration is implemented. ZTE can work with operators worldwide to build an open, win-win new integration service ecosystem.
A digital integration center is the capability support platform after the digital transformation of virtualized integration services. The key to AI-empowered integration services is to introduce AI capabilities into the integration center to build self-intelligence capabilities of NFV integration and promote the evolution of digital integration to digital intelligence.
ZTE's digital integration platform has built an integration capability center with the integration capability asset library (iCAL) as the core, and an integration project center with the integration software defined integration (iSDI) as the core. In the future, the platform will add the integration intelligent analysis engine (i2AE) and the integration intelligent analysis agent (i2AA), as shown in Fig. 1. The i2AE introduces AI capabilities, offers AI data analysis tools and training model libraries, and provides self-intelligence support for integration services based on data mining and deep learning of integrated digital assets fixed in the iCAL through algorithms such as policy control, machine learning, federal learning and intent-driven. The i2AA calls data between the AI-empowered digital integration platform module and the i2AE. As an interoperable interface for data interaction, the i2AA is also used to interconnect internal and external general AI platforms and to call external AI capabilities to continuously empower integration scenarios. In addition, the i2AA builds an intelligence center for digital integration based on the AI capabilities. Through the architecture of "capability center + project center + intelligence center", the digital integration platform will combine digital integration with AI technologies to build a ubiquitous NFV-integrated AI capability system and facilitate operators in their digital transformation.
Compared with application scenarios such as virtual assistant and machine processing automation, telecom operators' requirements for AI are still focused on efficiency improvement and costs reduction. Therefore, after the introduction of AI capabilities, the virtualized digital integration platform will adhere to digital and intelligent integration, focusing on the improvement of core technologies concerning system architecture and end-to-end self-evolution. The platform helps operators build a "smart brain" for integration services, which can better improve the efficiency of NFV network construction and reduce integration costs. As a result, operators can jointly build integration capabilities and tap the value of integrated data, so as to expand more business opportunities and drive changes in service model.