From AlphaGo and driverless cars to smart speakers and smart households, Artificial Intelligence (AI) technologies are gradually integrating with physical economies and industries. Smart products are constantly improving people’s lives and accelerating the construction of a smart society.
Our society will evolve to be ‘smart’ over the coming 20 to 30 years. Huawei envisions ‘bringing digital to every person, home, and organization for a fully connected, intelligent world.’ However, both challenges and opportunities will present themselves during this social transformation.
AI Growth Pains
As part of this AI trend, a full range of enterprises are rigorously researching this field; however, technology improvement comes with its own challenges, and we are facing AI development pain points that need to be quickly resolved.
First, AI needs to simulate human intelligence. We still do not fully understand how human intelligence is generated, and brain science research is still in the exploration stage.
The second pain point is computing capability. As evidenced by the use of large numbers of Central Processing Units (CPUs) and Graphics Processing Units (GPUs) to realize AlphaGo’s victory over 9-dan ranked Lee Sedol, large-scale computing platforms are often required to complete single, complex tasks, such as a game of Go. In light of this, what is the future of AI development?
The third pain point is data. We need to integrate AI with each industry, make systems generate a significant amount of data, and transmit the data to AI systems to make them smarter through continuous machine learning. However, the current industry environment cannot satisfy such conditions. Therefore, the data necessary to fully support AI development is currently insufficient.
Huawei’s Three Strategies for AI Development
To address these pain points, Huawei has adopted the following strategies for AI development: Technical innovation, collaborative business practices, and open cooperation.
- Technical Innovation: Building Full-stack Technical Capabilities
We conduct technical innovation in two aspects. First, we continue to increase our investment in R&D. Currently, Huawei invests in the R&D of real-time analysis and mining, in-depth user modeling and analysis, natural language processing, and analysis theory evolution technologies. The purpose is to gradually become familiar with basic theories and computing capabilities for further AI developments.
Second, we build full-stack technical capabilities using a ‘cloud-pipe-device collaboration’ architecture, where ‘cloud’ is enterprise intelligence, ‘pipe’ is the variety of equipment that connect clouds and devices, and ‘device’ is for the intelligent terminals. Huawei has formulated a strategy to support the development of AI technology that coordinates the interactions between next-generation platforms, applications, large-capacity information pipes, and a wide spectrum of intelligent devices. To be specific:
• Cloud: Huawei released the cloud Enterprise Intelligence (EI) solution in 2017 to provide one-stop platform services for business, industry, and government. The solution includes a heterogeneous computing platform and three types of enterprise-level intelligent cloud services — basic platform services, general services, and scenario-based solutions.
• Pipe: Based on AI analytics, Huawei builds an Intent-Driven Network (IDN) for customers to implement network self-adaptation, self-optimization, and automatic configuration.
• Device: In 2017, Huawei released Kirin 970, the world’s first smartphone chip with a dedicated Neural-network Processing Unit (NPU). The most significant highlight of the NPU module is the built-in Cambrian Period-1A GPU-based AI unit. The AI breakthroughs included with the Kirin 970 indicate an important watershed moment in the transition from traditional smartphones to all future phones that will include AI capabilities.
- Collaborative Business Practice: Huawei Provides Solutions Based on Customer Requirements
For business practices, we take two approaches. First, Huawei has added AI to our internal production, logistics, and after-sales services to improve operations management and support a sales volume of over USD 100 billion. For example, Huawei requires a large number of customs declaration orders every year to support global logistics operations. Previously, each manifest was processed manually with great inefficiency. AI technology has been applied in the form of optimization measures for the entire process: As intelligent documents are now automatically identified, entered, classified, and archived. This has increased the accuracy of shipment forecasts by 30 percent, decreased logistics expenses by 30 percent, and significantly improved efficiency by shortening the time to import customs declaration documents from 24 hours to one hour.
Second, Huawei joins hands with industry partners to provide customers with AI-based industry solutions. For example, Huawei and the Shenzhen Traffic Police Bureau have jointly innovated to build a ‘smart brain’ for Shenzhen’s roadway infrastructure. The system can detect the traffic information of each lane in real time through video, embedded magnetic coil, and microwave connected sources that provide traffic police with a complete picture in a timely manner. The result is an increase in status detection and reporting accuracy to 95 percent or higher. By applying AI technology, the efficiency needed to identify traffic violation snapshots has improved by 10 times; and a precise traffic-signal-control mode has been developed that, when combined with spatio-temporal analysis of traffic volume and roadway incidents in multiple dimensions, has improved road capacity by more than 8 percent.
- Open Cooperation: Building a Global AI Ecosystem
In the AI field, Huawei has adopted an open cooperation strategy that focuses on the collaboration with ecosystem partners to build a global ecosystem. For the implementation of this strategy, Huawei adheres to a policy of maintaining clear business boundaries for the purpose of developing a platform that supports other platforms and operates as the foundation of the larger ecosystem. Additionally, Huawei has created many joint innovation labs. Together, through joint solution development with partners, Huawei will continue to achieve business successes made greater by the combined strength of many parties. At present, Huawei runs 20 OpenLabs around the world and is working with more than 500 partners to incubate more than 200 joint AI solutions.
We believe that our heavy R&D investment and open cooperation strategy will strengthen our core competitiveness in the AI field and promote further AI development.