Since 1956, progress in Artificial Intelligence (AI) has ebbed and flowed, but recently, advances in big data and computing power have transformed AI into a technology with increasingly profound effects on our lives. Never before has AI impacted our everyday lives as it does today, but we can only effectively apply a technology after it is correctly understood and positioned. How, then, should we go about positioning AI?

HUAWEI CONNECT's theme in 2018 was "Activate Intelligence." It emphasized the technical innovations that are chiefly driving the growth in AI and its application throughout industry.

Xu Zhijun, Huawei's Rotating Chairperson, released Huawei's AI strategy, which focused on the current problems that AI can solve and the areas where AI can add the most value.

For AI, Huawei is investing in technology, talent, basic research, and industry applications, effectively building a full-stack solution, with an open ecosystem and talent cultivation to enhance solutions and improve internal efficiency.

In the past year, HUAWEI CLOUD has been integrating industry wisdom with AI in eight industries and over 200 projects by leveraging its technical innovations. With its Ascend series chips, Huawei can now deliver full-stack solutions that include chips, chip enablement, training and inference frameworks, and application enablement. Huawei offers deployment environments for a variety of scenarios, including the public cloud, private cloud, various edge computing scenarios, IoT industry terminals, and consumer terminals.

Intelligence Demanded by the Cloud2.0 Era

In the future, the cloud will be the principal environment for most enterprises to explore AI solutions due to the requirements for massive computing and storage resources.

Zhang Yuxin, CTO of the Huawei Cloud Business Unit, believed that the arrival of Cloud2.0 leads to several business scenario changes, as compared with Internet development in past decades.

Initially, in Cloud1.0, enterprises migrated their key applications to the cloud. The customer entity, entity demands, and service modes also changed with the advent of cloud computing. Government and enterprise markets were the first targets. According to Zhang, with the advent of Cloud2.0, "the main cloud applications are now switching from personal entertainment and consumption fields to the production field."

Second, traditional Internet applications have encountered bottlenecks. In the past, Internet applications mainly relied on traffic dividends. In the Cloud2.0 era, the core of new Internet business changes from traffic dividends to data dividends.

Huawei Vmall is one of the largest e-commerce companies in China, serving 400–500 million users worldwide. Its services are carried on HUAWEI CLOUD, which has become the choice for multiple Internet customers. The technical platform of HUAWEI CLOUD helps Internet customers improve performance and guarantee smooth processing, even during peak traffic periods. According to Zhang, "the Cloud1.0 era had distributed, automatic, large-scale, and elastic features, while the Cloud 2.0 era now requires AI, big data, and computing power, as well as a comprehensive system architecture for enterprise applications and Internet applications besides the basic demands for security and reliability."

Unique, Full-Stack Technology Innovation

Zhang believes that HUAWEI CLOUD offers many innovative and full-stack basic technologies, which continuously add AI value and overcome bottlenecks from traffic dividends to data dividends.

The majority of Chinese cloud vendors are not currently able to make chips. Huawei has released AI chips, and one Ascend chip is configured for high-specification H-series VMs. The Ascend 310 chip delivers computing performance that is 16 times faster than other chips.

For a range of AI computing scenarios, Huawei offers a variety of services covering VMs, containers, and bare metal servers. You can configure services with the highest 512 TOPS inference performance as required by using ECS and CCI. For training scenarios, Huawei provides ECSs and BMS with the computing capability of up to 2 PFLOPS.

Also at HUAWEI CONNECT 2018, Jia Yongli, the general manager of the EI Service Product Department of Huawei Cloud BU, announced that HUAWEI CLOUD EI will provide 45 services and 142 functions targeted toward improving basic computing capabilities to achieve inclusive AI covering all of the scenarios available for AI senior data scientists, data engineers, common IT engineers, service personnel, and others.

Huawei has developed the Hi1822 chip and launched the industry's first intelligent NIC with the single-chip rate of 100 Gbit/s, enabling Huawei's network rate to be the first to enter the 100G era. The 100 Gbit/s intelligent NIC also achieves a breakthrough in network forwarding. In addition, network connections among millions of nodes in data centers is enabled through reconstruction of the control-plane architecture.

For Huawei's cloud storage support of key enterprise applications, Huawei has adopted an innovative EC Turbo technology based on years of investment in mathematics and algorithms, effectively making Huawei the leader in both disk usage and valid capacity. Now, Huawei applies these HUAWEI CLOUD technologies to meet customers' requirements for data intelligence.

Huawei has made significant technical breakthroughs in vertical sectors to react to the challenges of providing enterprise intelligence. These advances include chips, hardware, data center management, basic cloud services, application development platforms, application deployment and O&M platforms, tool frameworks, basic AI models and algorithms, and domain-specific models and algorithms.

Driven by Technologies and Customer Requirements

Huawei has a three-stage strategy: collect, develop, plan for the future. The cloud field requires long-term investment in underlying technologies. Multiple technologies in research have not yet moved to the production environment.

Huawei always begins technical preparations 3–5 years, or even 5–10 years in advance, so they have invested nearly 400 billion RMB in R&D in the past eight years, with nearly 90 billion RMB in this past year alone.

Customer requirements and technologies drive continuous innovations in HUAWEI CLOUD. Today's rapid pace of development demands innovations in different areas. HUAWEI CLOUD employs a new method to continuously innovate with customers interactively through multiple iterations. Huawei provides customers with professional data engineers and scientists to aid them in abstracting business models, classifying and labeling data, and training models for trial use in their systems. Based on trials, engineers then tune and retrain the model to the necessary level of precision before integrating it into the customer's system.

At the keynotes on the second day of HUAWEI CONNECT 2018, HUAWEI CLOUD officially released the EI City Intelligent Twins solution, which is built on a digital twin and leverages multiple technologies such as AI, big data, edge computing, and IoT to deliver a complete system covering data generation, data analysis, and data closed-loop. The solution demonstrates how to control and manage aspects of the physical world more intelligently through digital computing power.

Li Qiang, technology division chief of the Shenzhen Traffic Police Bureau, said, "We are building a city traffic brain incorporating video cloud, big data, and AI core technologies and then constructing traffic intelligence through those technologies and policing to enable more intelligent cities. The Huawei cloud platform provides logical servers and complete deployment within seconds, significantly accelerating service deployment. With comprehensive traffic awareness and AI-assisted law enforcement, HUAWEI CLOUD supports high-quality, efficient transportation services for citizens."

A new era of AI has arrived. With powerful computing power and service upgrades brought by Huawei's latest AI chips, HUAWEI CLOUD remains at the cutting edge of innovation, accelerating value creation.