Getty Images/iStockphoto

Can Huawei untangle the AI mess?

Huawei is building a digital platform with “device-cloud synergy” to support artificial intelligence applications, but enterprises should be wary of being locked into a single AI stack

As with any emerging IT segment, the marketplace for artificial intelligence (AI) is messy for now.

While enterprises already have a variety of AI-related offerings to choose from – including AI-optimised hardware from Lenovo, cloud-based AI services from Amazon Web Services (AWS), and software from niche players such as Adatos – they have to architect an entire technology stack to run their AI applications.

As such, most AI initiatives are siloed, fragmented, and are not as pervasive across the enterprise as they should be. To address this gap, more technology suppliers are now looking to provide a single platform that powers all the AI that an enterprise needs.

Huawei has become the latest supplier to signal its intention to develop such a platform. At its recent analyst summit in Shenzhen, it said it wants to build a digital platform with “device-cloud synergy” to support AI applications across the enterprise.

Such a platform plays to Huawei’s strengths. The company already makes the AI-optimised Kirin smartphone processor with a neural processing unit, and it also offers cloud services, as well as networking, storage and server hardware for enterprises.

As for the software smarts, Huawei will rely on its partners, such as Chinese AI software specialist Yitu, to provide AI applications and algorithms for its enterprise customers.

The company was tight-lipped about whether it would develop its own algorithms for enterprise applications at some point, or extend the use of Kirin chips to edge computing devices that process data at the edge of the network in internet of things (IoT) applications.

It also declined to say if it would provide cloud services that enable businesses to take advantage of existing machine learning frameworks such as TensorFlow, claiming that it will make more AI-specific announcements at its Huawei Connect event later this year in Shanghai.

That said, Huawei isn’t about to cede all the AI-related software smarts to others. It is already planning to infuse AI into its carrier network offerings, helping carriers to improve network management through autonomous network capabilities.

Internally, Huawei is also building up its AI expertise. At its warehouses, for example, it claims to have optimised picking routes using AI, and increased efficiency by more than 30%. It has also improved the accuracy of its stock estimates from 30% to 80%.

Read more about AI in APAC

Eric Xu, Huawei’s rotating chairman, said these internal AI efforts will be incorporated into its AI offerings and made available to enterprise customers at some point. “We have been pushing the application of AI technology in every part of our internal operations to improve efficiency,” he said.

Such capabilities should make Huawei the supplier to watch in the years ahead as it builds up its AI platform that it hopes will increase adoption of AI among enterprises. According to Huawei, just 5% of enterprises are experimenting with AI, while less than 1% are using the technology in a meaningful manner.

Although Huawei’s moves are commendable and should help ease the infrastructure complexity of AI deployments, any pitch to adopt a full technology stack from a single supplier should be met with caution.

Unless Huawei’s platform is an open one that lets enterprises plug AI applications in and out easily, there is always the risk of being locked into a single platform – certainly not at a time when AI innovations are coming from nearly all quarters in the technology industry.

This was last published in April 2018

Read more on Artificial intelligence, automation and robotics

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.

-ADS BY GOOGLE

SearchCIO

SearchSecurity

SearchNetworking

SearchDataCenter

SearchDataManagement

Close