Huawei's Zheng Yelai on Powering AI

Sep 17, 2018

 [Shanghai, China, 17 September 2018] This is the age of transformation. We are all witnesses to how AI is changing the landscape and becoming pervasive in everything from voice assistance to facial recognition and a slew of other applications. The explosion shows no signs of letting up as everyone is talking about what the new accelerator is able to do in their respective industries and occupations. Specialized hardware platforms are the future for the tooling that make up AI applications - a direction being talked about at some of the biggest "tech coffee talk" venues like the World Artificial Intelligence Conference that opened in Shanghai. Zheng Yelai, Vice President of Huawei and President of Huawei Cloud BU, shared his thoughts on AI developments. 

The following is contents of the speech delivered by Zheng Yelai at the conference. 

AI will change every industry, occupation, organization, family, and individual.

AI is a collection of new general-purpose technologies (GPTs), including natural language processing and image recognition. If information technology brought about efficiency improvements, then AI brings significant change to production with much improved cost efficiencies. Over the past few years, Huawei has been exploring ways to apply the new set of tech to a wide range of industries, yielding some rather remarkable socioeconomic benefits. For example, in the transportation field, we built an intelligence-enabled urban traffic system in Shenzhen Municipality just over the border from Hong Kong. We turned this city that was amongst the most gridlocked metropolis anywhere into one with well-flowing traffic arteries. In the logistics industry, we have worked with customers to explore improvements in route planning, identification and sorting of hazardous and explosive materials, tamper-proofing cargo efforts, and a long list of other advancements. 

AI will change every industry, occupation, organization, family, and individual. We have been saying it is about finding ways to apply AI in specific ways to each industry, not just forcing AI into the industry structure. The talent structure of the organization today generally has business leaders at the top, senior managers/experts below them then the junior-level managers/operational backbone followed by entry-level employees at the bottom. Applying AI will change the talent structure. AI engineers will work in between each level and across the entire spectrum, from the business leaders at the top to the managers at the middle and the entry-level employees at the bottom, so all the repetitive work in each tier can be replaced with simulated intelligence utilities.

The change in the next two to three years may very well achieve this level of automation and acceleration. Gartner predicts that many industries and professions will change in huge ways before 2020.

AI is a new high point in IT-enablement

The application of GPTs will have a process to follow, of course. At Huawei, we generally divide industry adoption of tech into four phases. The first is exploration of technologies and applications to suit the landscape. The second stage is the mixing of technological development with the socioeconomic environment. The third stage is the equal promotion of technological development with the socioeconomic benefit. The fourth stage isstable development of tech to become general purpose.

We believe that the AI of today is at the beginning of the second phase. The developments are creating plenty of excitement along with the anxiety and confusion that comes with anything 'new'. One thing is certain though, AI is changing these industries and the impact will only deepen. That being said, we need to define our point of view. First, we do not believe that AI can be applied to or fix everything. Any technology that can be used to solve certain problems will not be able to solve others. We should focus on the problems that can be solved by AI and the fields that create value instead of focusing on problems that cannot be solved or areas that do not create value. Machines are expected to perform as well or even better than humans in more and more tasks as time goes on.

We also hold another important viewpoint: Choosing the right problem is more important than looking for a novel solution. Machine learning is more effective than humans for some patterns that can be followed and supported by data. One of our guest speakers shared on the core issue of data support this morning, which involves many issues, including who has sovereignty over the data and how to protect the privacy of the end customer. These issues urgently await resolution. AI is the new high point in IT-enablement because it is able to reduce production costs. It is a science combined with a set of compute technologies. It gets its inspiration from people yet it is much different from people. Like humans, AI is capable of sensing, learning, reasoning, and taking action. It can also do things better that previous IT enablement could not accomplish. The IT adoptions in the past helped improve human efficiency and reduced labor costs, but they could not reduce production costs through automated means as effectively as AI.  While AI can be used to replace repetitive human actions, it still needs specially trained engineers to tune the learning models and machines. So, where can such talent be found? AI engineers will include data scientists, subject-matter experts, and other professionals to fill the shoes. AI can even be used to solve the shortage in qualified personnel to run the apparatus. In addition to obtaining data scientists from outside the company, we are strengthening internal training and creating our own data scientists and using AI to develop the needed capabilities. This way, we can form the needed capabilities more quickly with our own talent base from which to draw. Again, choosing the right problem is more important than looking for a novel solution. This mindset applies here as well. We identify the problem then apply the right tech and attract, train, and develop the talent to help solve the issues and keep the model tuned. Using AI to solve the shortage of AI professionals means we need to focus on intelligent, automated, and easy-to-use tooling and platform services so that all enterprises can use the new pillar tech in a faster and more efficient manner to solve problems and create value. More adoptions means AI can become a common technology sooner – a general-purpose yet cutting edge technology that is inexpensive to apply. We call this general mindset "Inclusive AI" because it seeks to include everyone that can benefit. 

Powering AI Depends on Easy to Access, Use, and Deploy Compute Resources

The first problem that must be addresses is the high costs. The training processes involved with AI are extremely costly. Facial recognition, comprehensive traffic control, automatic driving, and other applications all require large outlays for the training model. 

The second problem is the difficulties in use. Lack of a unified development framework that can adapt to various application scenarios like training and reasoning and one that can work across public/private cloud, edges, and terminals adds to the complexities in adoption. The huge workloads involved in development, optimization, and deployment also complicate implementation.

The third problem is the difficulty in obtaining the needed compute power to run the tool house. The time needed to deliver GPU resources is lengthy and supply is limited. 

Huawei is working hard to find ways to reduce costs, simplify use, and make more resources available at the right price point. Huawei is releasing its AI strategy at HUAWEI CONNECT being held from October 10 to 12, 2018 in Shanghai. Full-stack, full-scenario solutions will also be on display. Huawei is building up an inclusive ecosystem and supporting research efforts at many high-end universities so enterprises can easily obtain and apply AI in their platforms. Huawei is also making the needed tooling available to AI developers so more applications can be developed to the benefit of all. All these efforts culminate into the Inclusive AI strategy from Huawei so everyone can get what they need, make easy use of cutting-edge tech, and rest assured that their profile remains fully secure throughout the entire lifecycle.