Oct 30, 2020| JD Technology

JD and Shanghai Municipal Government to Issue RMB 2B Yuan Coupons

Ling Cao

by Robin Luo

JD has joined hands with Shanghai’s municipal government to issue coupons worth RMB 2 billion yuan for local consumers during this year’s Singles Day Grand Promotion.

Consumers can enjoy discounts on JD using these coupons together with subsidies and coupons previously issued by JD. As China International Import Expo is around the corner, JD’s Singles Day promotion aims to boost strong and resilient consumption growth in Shanghai.

Starting from Nov. 3^rd, consumers who purchase the designated products, such as cellphones, smart home appliances and digital products on JD can enjoy price cuts as high as RMB 400 yuan thanks to the coupons. It is expected to benefit tens of millions of Shanghai residents. The coupons will be issued on Nov. 3^rd and Nov. 9^th.

This year, during the Singles Day, JD will offer over 200 million products at 50% discounts, and over 300 million new products to Chinese consumers.

(press@jd.com)

Mar 20, 2023| JD Technology

JD.com Recognized with China’s Most Prestigious AI Awards

Yuchuan Wang

by Yuchuan Wang

On March 14, JD.com received the Wu Wenjun Artificial Intelligence Science and Technology Progress Award, the most prestigious honor in China for work in artificial intelligence. In addition, Dr. Xiaodong He, vice president of JD.com, an Institute of Electrical and Electronics Engineers (IEEE) Fellow and a specialist in conversational AI technology, was given the Outstanding Contribution Award. Wu Wenjun (1919-2017) was a prestigious mathematician who obtained the highest national science and technology award in 2000.

As one of the founders of early multimodal AI, Dr. He was selected as an IEEE fellow in 2019 for his contributions to multimodal signal processing in human language and vision technologies. Under his leadership, JD.com launched the Yanxi AI application platform and a series of industrial solutions in 2018. Yanxi provides consumers and merchants 24/7 services, including shopping guidance and across shopping nodes– pre-sale, sale, and after-sales.

Unlike a common chatting platform, task-based dialogue systems aim to solve complex tasks happening in real life and these systems are crucial to user experience. For example, in e-commerce, customers have an expectation that the chatbot can perfectly solve problems such as aftersales consulting, price guarantee, transaction, payment, delivery service, product return and more. JD’s task-based system seeks to provide that service for the consumer.

Built upon the company’s experience in intelligent dialogue accumulated over the past 10 years, coupled with years of practice in various businesses across retail, logistics, finance, health, and more, Dr. He’s team has constructed a knowledge graph that is made of a four-layer knowledge system, 40+ independent systems, 3,000+ intentions and 30 million high-quality QAs that cover over 10 million retail commodities on JD.com.

Currently, Yanxi serves over 580 million customers, hundreds of thousands of merchants, 100+ municipal institutions, 880 financial institutions, 1821 large enterprises and nearly 2 million SMEs.

(yuchuan.wang@jd.com)

Feb 15, 2023| JD Technology

JD Cloud Launches ChatJD to Integrate ChatGPT into Technology Products and Services

Yuchuan Wang

by Tong Shen

JD Cloud announced this week that it had launched “ChatJD,” an industrial version of ChatGPT focused on the vertical industrial fields of retail and finance.

“JD.com offers the best scenarios to accelerate the scale implementation of AIGC & ChatGPT,” said Dr. Xiaodong He, vice president of JD.com, an IEEE Fellow, and a specialist in conversational AI technology.

The industrial ChatJD will roll out through one platform, two fields and five applications. It will build a human-computer dialogue platform that understands and generates tasks in natural language processing, with parameters estimated to total hundreds of billions.

Built upon the company’s experience in intelligent dialogue accumulated over the past 10 years, coupled with years of practice in various businesses across retail, logistics, finance, health, and more, ChatJD is more focused on retail and finance. It has constructed a knowledge graph that is made of a four-layer knowledge system, 40+ independent systems, 3,000+ intentions and 30 million high-quality QAs that cover over 10 million retail commodities on JD.com.

Additionally, ChatJD will develop five application scenarios, ranging from content generation and human-computer dialogue to user intent understanding, information extraction, and sentiment classification, which will play an important role in areas such as automated customer consultation and service, smart marketing content generation, financial research report and more.

Since 2019, JD.com has begun to explore NLG (natural language generation) to research in AI-generated marketing content in house, such as generating product titles, selling points, and scripts for livestreaming and more. The service has covered more than 2,000 product categories and has been applied across multiple marketing channels on JD’s app, generating a total of more than 3 billion words of text. In addition, developed from JD’s early smart customer service chatbot, Yanxi AI platform provides consumers and merchants 24/7 services in shopping guidance and many others, covering end-to-end shopping nodes including pre-sale, sale, and after-sales. Yanxi is currently used by 588 million consumers and 178,000 merchants on JD.

Dr. He believes the impact of ChatGPT to the industry may rival what iPhone has produced. ChatGPT’s biggest innovation lies in the generation of natural language, which gradually sorts out the user’s intention through interactive dialogue and generates intact texts. From the technical point of view, ChatGPT innovatively uses semi-supervised learning algorithms compared to the previous unsupervised deep learning algorithms, and makes the model for content generation more organized. To some extent, it is actually a course correction to the past obsessive pursuit of mass (parameters) and unsupervised learning.

(press@jd.com)