As a General-Purpose Technology (GPT), Artificial Intelligence (AI) has multiple uses and creates huge ripples across society. AI will be the new engine that drives human society forward, just as the invention of the wheel was for human civilization so long ago. For example, AI is widely used in many industries today, such as the Internet, public security, education, healthcare, and transportation. Stakeholders are also looking to make AI-driven breakthroughs in industries such as finance, manufacturing, energy, materials, and retail.
However, the wheel was not invented in a single day, which is also true of AI. Although AI has made many breakthroughs, and will continue to do so, there are still some obstacles to overcome. We need to make material changes in model algorithms, compute power, data security, talent cultivation, and more to make AI universally accessible to every aspect of society. The 10 changes listed below (Figure 1) are the basis for future AI development.
Huawei understands these changes as both the future development trends for the AI industry and a strong basis for its own AI strategy. Huawei’s strategy consists of five key aspects: Invest in basic research, build full-stack solutions, invest in an open ecosystem to cultivate talent, enhance existing solutions, and improve internal efficiency.
Figure 1. 10 changes required to make AI more accessible
Huawei’s All-Scenario AI Chipset and Intelligent Computing Platform
We are entering a thrilling era of intelligence where things can sense and connect to one other. Breakthroughs in Information and Communications Technology (ICT) make our society more intelligent but require increased computing power. This poses a great challenge for current AI computing capabilities. In view of this, Huawei has launched the Ascend series of chips — the world’s first all-scenario AI chipset — and the Atlas intelligent computing platform powered by Ascend.
The Ascend 310 chipset is a heterogeneous AI System on a Chip (SoC) that is designed on Huawei’s proprietary Da Vinci architecture. It boasts unrivaled energy efficiency, with 16 TeraOPS (TOPS) @ INT8 computing on a single chip. It comes with a 16-channel HD video decoder and consumes less than 8 watts. The Ascend 310 is perfectly suited for devices deployed on the terminal and the edge.
Huawei has launched a series of AI hardware platforms based on the powerful Ascend 310 chip for terminal-side and edge computing scenarios (Figure 2). Available in various product form factors, such as accelerator modules, cards, AI edge stations, and appliances, the company’s all-scenario AI infrastructure solution can be widely used in various fields for Device-Edge-Cloud scenarios. As an integral part of Huawei’s Full-Stack AI Solution, the Atlas intelligent computing platform provides supreme compute power to help customers embrace an AI-fueled future. It is backed by a full-stack software layer and supports most mainstream frameworks. It also provides easy-to-use code porting and model conversion tools to facilitate application development and deployment.
- Atlas 200 AI accelerator module: Packaged in a form factor half the size of a credit card, the Atlas 200 supports 16-channel, real-time HD video analytics. When deployed on devices such as cameras and drones, it consumes only about 10 watts of power.
- Atlas 300 AI accelerator card: The size of a standard half-height, half-length Peripheral Component Interconnect Express (PCIe) card, the Atlas 300 is designed for data center and edge server scenarios. It supports multiple data precisions. A single card provides 64 TOPS @ INT8 computing performance, which infuses superb compute power into deep learning and inference applications.
- Atlas 500 AI Edge Station: An industry-leading edge product for AI processing, the Atlas 500 processes up to 16 channels of HD video while taking up no more room than a Set-Top-Box (STB). The AI edge station is well suited for a broad range of applications, such as transportation, healthcare, unattended retail, and smart manufacturing.
Figure 2. Atlas product portfolio
Open Ecosystem Investment Cultivates Talent
At Huawei Connect 2018, Rotating and Acting CEO of Huawei Eric Xu unveiled Huawei’s AI strategy and a full-stack, all-scenario AI solution. Building on its leading-edge AI platform, Huawei will continue to join with various industries to tap into new opportunities and to drive industry transformation through ubiquitous intelligence, leading the way toward a fully connected, intelligent world.
To make this vision a reality, Huawei plans to leverage its full-stack intelligent computing platform and build an industry ecosystem with the end customers in mind. Huawei will pursue a strategy of pairing AI applications with Independent Software Vendors (ISVs) to improve industry efficiency. Huawei is dedicated to building an open AI ecosystem based on the following four aspects:
- ISV ecosystem:Develop complete, efficient, easy-to-use, and low-cost solutions for various scenarios in different industries.
- College ecosystem:Train future developers and enrich the industry talent pool. By developing AI courses, building laboratories, publishing books, training teachers, and holding student contests with universities, Huawei’s AI technology has become the starting point for many future developers. Many universities in China have collaborated with Huawei to develop courses and to write text books. The industry, academy, and research communities have celebrated these efforts and achievements. In December 2018, Huawei joined the School of Electronics Engineering and Computer Science of Peking University to hold a 100-School Teacher AI Summit at the Yingjie Exchange Center of Peking University. More than 340 teachers from 157 universities and research institutes attended this event. The summit explored colleges’ and universities’ efforts to cultivate AI researchers and pioneers.
- Entrepreneur ecosystem:China has four of the top 10 cities in terms of global venture capital investment. Most notably, Beijing has surpassed California’s Silicon Valley and New York as the world’s top venture capital hub. Over the next few years, China’s environment of innovation is expected to produce the next generation of startups that will reshape the world. In view of this, Huawei will strive to help future ISV partners accelerate their growth.
- Open-source ecosystem:For the ICT industry, the significance of open source cannot be overstated. In terms of AI development, Huawei will embrace open source, including support for mainstream open-source AI frameworks, and proactively open up its AI technologies to the community for secondary development.
An Open Ecosystem for an Intelligent World
Huawei will work with customers in an inclusive, sustainable, and future-facing spirit to drive intelligent enterprise digital transformation by leveraging the power of the open ecosystem. Currently, dozens of industry partners have discussed collaborating with Huawei to use the Atlas intelligent computing platform. Related solutions and application scenarios cover various industries, such as public utilities, the Internet, smart transportation, energy, and retail. Huawei looks forward to building an open and robust ecosystem with more partners in the future.