President of JD AI Bowen Zhou discusses the need for collaboration as a way forward for smart retail and AI
By Min Chen
With China’s tech giants embracing smart retail in recent years, it’s safe to say that a revolution in the space has arrived. While rivalry may typically characterize such a narrative, technologists and business leaders alike have pushed for a more collaborative course toward the country’s vision for the future of retail.
One proponent of this approach is Vice President of JD.com and President of JD Artificial Intelligence (JD AI) Bowen Zhou, who Jumpstart had the opportunity to interview before his fireside chat at RISEConf 2019, ‘The Promise of AI: From retail to the rest of the world.’
Zhou is a renowned computer scientist who has published over 100 articles on topics such as deep learning, automatic speech recognition, and natural language processing. With the intent to advance his research to impact the real world, he joined IBM in 2003 and took on the roles of IBM Watson Group Chief Scientist and Director of the AI Foundations Lab of IBM Research before being appointed to his current position at JD.com in 2017.
He was attracted to China’s AI sector because he considers it to be “a center of gravity” for both the development of core technologies, and solutions that lower cost and improve current applications.
“China is unmatched with its rich dataset and it’s an environment that can stimulate innovations to benefit from technologies such as AI–that’s a key reason that draws me to this amazing journey,” says Zhou.
JD.com is a dominant player in the country’s tech landscape and is China’s largest retailer, and the third largest Internet company in the world after Amazon and Alphabet (JD.com). With an asset-driven business model–where the company manages purchasing, storage, fulfillment, and a direct sales channel–JD.com has seen tremendous growth since its ecommerce platform launched in 2004.
It reported RMB462 billion (US$67.2 billion) in net revenue in 2018 and saw the number of user accounts reach a jaw-dropping 320 million as of the second quarter of this year, or around 40% of all Internet users in China (JD.com). For JD AI, the aim is to deploy such resources to empower every aspect of retail through integration.
“When you consider the value of AI, you need to go beyond a single point,” says Zhou. “What is the entire lifecycle of the industry, and how can we design from the top?”
He explains that this design approach entails connecting all of the isolated AI applications in the company’s infrastructure–such as enhancing offline shopping experiences, supply chain, and customer feedback–to fully digitize the retail process. Ultimately, the data will be shared with upstream retail operations, such as manufacturing, R&D, design, marketing, and logistics.
“This is what we call ‘from customer to manufacturer.’ That’s a goal we have at JD AI,” adds Zhou.
JD AI is moving quickly in making this goal a reality. Notably, it has partnered with Fung Retailing Group to launch Hong Kong’s first AI-powered checkout at two Circle K shops. The computer vision-based system is capable of facilitating a four-second checkout time by recognizing up to five items per second with an accuracy rate of 97%.
“We’ve deployed this for three months now. So far, the system hasn’t misrecognized anything and more importantly, the speed of checkout time is up by 30%,” adds Zhou.
He anticipates this type of checkout station to be launched at scale in a short amount of time, especially in cities like Hong Kong, where long queues can lead to revenue loss. The collaboration is the first part of an end-to-end system that will manage products, pricing, storage, and other smart retail experiences for the two companies.
For Zhou and his team, the mission is to not only accelerate retail innovation within JD.com but to do so for smart retail in China as a whole. The challenge they’re facing is tempered growth resulting from the general maturation of the innovation sector over the past decade. Having a great idea is no longer enough, and companies have to outperform in every way to break through the noise.
“If you want to succeed in applying technology such as AI, you need to fundamentally think about: does this save cost, does this improve efficiency, and does this enhance the user’s experience? You’ve got to have one of them; if you have all three of them, fantastic. But if you have none of them, then your product is probably premature,” says Zhou.
He adds that having a long-term view of the industry is also key to creating value, which requires a visionary leader and resource-intensive algorithmic and statistical research. For startups, Zhou advises that they focus on vision, as they may be lacking in resources. To help fill this gap for some startups, JD AI launched a global AI Accelerator last year, where it will provide technical guidance–and in many cases, funding–for accepted companies in addition to training and mentorship.
“The companies I’m looking for have a clear idea about what problems they’re addressing. As long as they are clear about that, we can help them with our technologies,” says Zhou.
An example is FaGouGou, an AI-powered legal consultancy that was a part of the program’s first cohort last year. JD.com provided the startup with access to its speech recognition technology, AI algorithms, and legal data to aid the development of a chatbot. This proof-of-concept has already been rolled out to many of JD.com’s suppliers, partners, and merchants.
Although the role that startups will play in China’s retail revolution looks promising, a shortage of tech talent in the country is another obstacle to growth. Zhou believes the Accelerator is also the answer to overcoming this scarcity.
“There are countless ways people can benefit from AI, but to make this happen, we need talents who understand AI and have industry know-how,” he adds. “That’s why it’s very important for us to build and empower a community and an ecosystem.”
Looking ahead and beyond retail, Zhou is most excited by advancements in AI that brings its behavior closer to that of humans. He explains that intelligence and language develop hand-in-hand, as humans learn through conversation and reading. If a machine is also able to internalize knowledge by reading and interacting with humans and other machines, then they would be able to draw conclusions without copious amounts of data.
“The most exciting would be combining reinforcement learning and natural language understanding capabilities, so a machine doesn’t have to be trained with tons of data to recognize a cat from a dog,” says Zhou.
From an application perspective, he’s fascinated by the intersection of AI and other cutting edge technologies such as 5G. Zhou uses the example of remote surgery to explain how AI can synergize with 5G’s high density and low latency functionalities; AI imaging can be used to analyze the patient’s condition in real-time to help surgeons make decisions on the fly. By combining 5G and AI, JD is currently researching how to create customized broadcasting using these two technologies in anticipation of the 2022 Winter Olympics in Beijing.
“The audience sees the same video, but if you are a proficient skier, you would want to see detail in the technique. And if you are a beginner, you want to see more of the playback. Combining AI with 5G will essentially enable a personal director working your TV, where it can synthesize, generate, combine video content personalized for you with no latency,” says Zhou.
While it’s difficult to predict what’s to come in 50 years for a technology as far-reaching as AI, Zhou does leave us with the idea that “there’s a tendency for people to overestimate what we can in the short-term and underestimate what we could do in the long-term.”
We’ll take that as a cue to let our imaginations run wild.