besjunior -

Huawei Connect 2018: Huawei tackles barriers to effective AI

At its annual customer and technology event, Huawei chairman Eric Xu set out 10 key challenges facing AI, and revealed how the networking supplier’s strategy will help to address these

Effective use of artificial intelligence (AI) in both the enterprise and consumer spheres will depend on overcoming a number of barriers in areas such as computing power, automation and skills, according to Huawei’s rotating chairman Eric Xu.

In a keynote address delivered at Huawei’s annual customer and technology event in Shanghai, Huawei Connect, Xu said that after a number of false starts over the years, the future of AI is clearly now a highly dynamic one, so it is now more important than ever that the technology is properly defined and managed for its true value to be unleashed.

“AI will help us to find more efficient solutions to problems we already know how to fix. It will also help us address problems that, to date, have remained unsolved,” said Xu. “As companies, if we want to stay ahead we need to adopt an AI mindset and use AI concepts and technologies to tackle existing and future problems.

“Nevertheless it is important to keep in mind that AI is not a cure all. No technology can solve every problem. We need to focus on areas where AI can create the most value, not on problems that AI isn’t equipped to solve. Finding the right problem is more important than devising a novel solution.”

AI is surging ahead in development, with more than 20,000 academic papers on the subject published in 2017, 22 countries having announced national AI plans, thousands of new AI startups, and venture capital (VC) investments topping $14bn per annum. But Xu pointed out there is a growing gap between its development and sluggish enterprise adoption – only around 4% of enterprises have invested in or deployed the technology.

He set out a number of changes that need to take place in the IT sector for these gaps to close and for the potential of AI to be realised. At the most fundamental level, it will require a massive growth of computer processing power, which he said urgently needs to become more abundant and affordable.

Looking beyond that, the growth of AI is exposing multiple weaknesses in existing algorithms, so future algorithms will need to be more efficient, secure and explainable.

AI will also need to synergise better with other technologies, such as cloud, the internet of things (IoT), edge computing, blockchain and big data. Particular attention will also need to be paid to the gap between how AI performs in test environments and real-world applications.

In human terms, the future of AI will hinge on automation. Right now, said Xu, AI projects are quite labour intensive, but in the future they will need to automate aspects such as data labelling and collection, feature extraction, model design and training.

This will also require a readjustment in the IT skills base (as already explored at Huawei Connect 2018), said Xu. “At present, AI is a job that can only be done by highly skilled experts. There aren’t enough mature, stable and extensive automation tools,” he said.

“Moving forward, we need a one-stop platform that provides the necessary automation tools, making it easier and faster to develop AI applications. When this platform is in place, AI will become a basic skill of all application developers, even all IT workers.”

Automation will also help overcome the scarcity of data scientists, which is an obstacle to AI progress. Xu said addressing this challenge would require an “AI mindset”, meaning providing intelligent, automated and easy-to-use AI platforms, tools, services and education to foster a huge number of data science engineers.

Huawei details AI strategy, full-stack AI portfolio

The challenges facing AI have inspired Huawei’s own AI strategy, said Xu, which hinges on five priorities: investing in AI research to develop fundamental capabilities; building a full-stack AI portfolio; developing an open ecosystem and talent; strengthening its existing technology portfolio; and driving operational efficiency within Huawei’s own business.

Xu went on to launch Huawei’s full-stack “all-scenario” AI portfolio. The stack is composed of four layers, the first of which comprises one or the other of Huawei’s first two AI chipsets.

The first chipset is the Ascend 910, which Xu claimed had the greatest computing power density ever achieved on a single chip. The second, the Ascend 310, will run low-power computing for smartphones and IoT devices, among other things.

The second layer, Compute Architecture for Neural Networks (Cann) will offer a chip operators’ library and automated operators’ development toolkit.

The third layer, called MindSpore, is a unified AI training and inference framework for devices, edge computing and cloud.

The fourth and final layer centres on application enablement, and comprises a set of full-pipeline services, known as ModelArts, hierarchical APIs, and pre-integrated solutions.

“All scenarios means that Huawei is able to deliver pervasive intelligence for a fully connected intelligent world,” said Xu. “Full stack means that Huawei is able to provide AI application developers with unparalleled computing power and a strong development platform.

“We are working towards making AI more inclusive, making it affordable, effective and reliable for all.”

Read more about AI

Read more on Artificial intelligence, automation and robotics

Data Center
Data Management