JD.com unveiled its own 100-billion-parameter large language model (LLM) called ChatRhino, or Yanxi in Chinese, at the 2023 JDDiscovery tech summit earlier this month. 

The AI model – which is the same type of technology used to train chatbots like ChatGPT – is designed to address challenges in sectors such as retail, logistics, finance, health, and urban planning, by leveraging a mix of 70 per cent generalised data and 30 per cent native intelligent supply chain data. 

JD.com has also unveiled plans to develop a generative large model that would enable businesses to use ChatRhino in real-world scenarios, and eventually opening it up for commercial applications in early 2024.

The story so far

According to the chair of JD.com’s technology committee Peng Cao, ChatRhino  will provide an unmatched, bespoke solution for the retail industry by leveraging a unique blend of generalised data and industry-specific, intelligent supply chain data. 

“By capitalising on this extensive, rich data network, ChatRhino presents a more nuanced and efficient AI solution for industry-specific challenges compared to other AI models,” he told Inside Retail.

He went on to say that JD.com has a complex and rich source of supply chain data, encompassing over 10 million proprietary e-commerce SKUs and 50 million industrial product SKUs while serving eight million enterprise clients and over 2,000 industry belts in China.

The ever-evolving retail landscape

Peng said that given the ever-changing retail landscape, ChatRhino’s strength lies in its robust industry generalisation capabilities. 

“JD.com’s vast operations across sectors including e-commerce, logistics, technology, and telemedicine, allow ChatRhino to continually learn from billions of high-quality, interactive data points,” he added.

Furthermore, JD.com’s reach extends to nearly 600 million active users and encompasses a broad range of offline operations, such as supermarkets and electronics stores. By utilising this comprehensive, multi-sector data, ChatRhino stays responsive and relevant to an evolving retail industry.

Its desktop-as-a-service cloud computing offering, Youjia, which delivers virtual desktops to retail end users over the internet, now has multi-dimensional data not only from the enterprise, but also from JD’s larger ecosystem.

“It’s easier to use, lowering the learning curve and enhancing the interactivity while simplifying the interaction between data platforms for a seamless user experience,” he noted.

According to Peng, ChatRhino now offers a variety of interactive channels, including intelligent customer service, live streaming, mobile phones, and other smart terminals. 

He said these upgrades offer more engaging, flexible, and user-friendly interactive experiences.

E-commerce marketing

He also explained that JD.com’s AI-generated content (AIGC) marketing platform integrates product data and the capabilities of ChatRhino to derive a deep understanding of product characteristics. 

In turn, it elevates efficiency and quality of marketing for merchants, particularly SMEs. With a single product image, the platform can generate an extensive range of essential e-commerce visuals, including marketing posters and product images.

The platform has the ability to reduce the cost of each visual set by 90 per cent and cut production time down from seven days to just half a day. One of the initial beneficiaries of the AIGC platform was JD’s private label brand J.Zao.

For retailers and merchants, JD has launched a multimodal digital presenter solution for applications such as livestreaming and shopping guidance.

To date, JD’s multimodal digital presenter has been integrated into over 4,000 brand and merchant livestreaming rooms, generating over RMB 800 million yuan in GMV.

Cost-effectiveness is key

Peng reiterated that ChatRhino contributes significantly to cost-effectiveness, which is a cornerstone of JD’s business philosophy.

“With this platform, users can effortlessly strategise, make personnel arrangements, set up campaign pages, and even bulk-create marketing copy and materials,” he stressed.

This reduces human-computer interactions from over 2,000 to under 50, paving the way for a hundredfold increase in marketing production efficiency.

In addition to general data, Peng said ChatRhino employs a wealth of high-quality data resources, spanning JD’s e-commerce and retail sectors, which also enhances the efficiency and reliability of supply chain management.

He added that the company’s LLM development is deeply interconnected with its industry partners, forming a symbiotic relationship that encourages innovation. 

From optimising supply chains to enhancing customer experience and improving marketing efficiency, Peng reaffirmed JD.com’s aim to leverage ChatRhino and other AI solutions to foster growth, efficiency, and transformation in the retail sector.

“Future collaborations may also include exploring areas such as supply chain optimization, customer experience, marketing efficiency improvement, and intelligent investment advisory,” he elaborated.

In a related development, Canadian e-commerce firm Shopify announced that it will be launching an artificial intelligence assistant for merchants on its platform. The assistant called ‘Sidekick” will be embedded as a button on Shopify and answer merchant queries.