Xin Lang Cai Jing·2025-09-28 05:56Core Insights - The integration of AI into e-commerce is accelerating, with companies like JD.com and Taobao focusing on deep integration of large models into their business operations [1][2] - JD.com has introduced its e-commerce AI architecture, Oxygen, and plans to launch more intelligent features in its app, including smart shopping guides and 3D product presentations [1] - Various AI models have been developed based on JD.com's JoyAI model, serving as the "brain" and decision-making center for intelligent e-commerce [1] Group 1: JD.com's AI Initiatives - JD.com is developing functional intelligent agents for product display, procurement, customer service, and marketing to enhance the smart e-commerce platform [1] - The company has opened its multi-agent collaboration framework, OxyGent, and the Oxygent-9N-xLLM model inference engine for future technological openness [1] - JD.com is also focusing on enhancing product presentation technology, particularly in XR and 3D fields, to create virtual try-ons and 3D live streaming [2] Group 2: Industry Trends and Competitors - Other e-commerce platforms like Taobao and Meituan are also accelerating their AI initiatives, with Taobao launching features like "AI Assistant" and AI price comparison tools [2] - Meituan has introduced an independent C-end intelligent agent, "Xiao Mei," focusing on local life services and providing intelligent interaction based on user preferences [2] - The AI e-commerce sector is still in its early stages, with companies exploring different strategies and competitive advantages based on their unique business models [3] Group 3: Challenges and Future Directions - The main challenge in scaling 3D content creation lies in the time and cost involved in producing 3D models, which can take months and be expensive [2] - The phenomenon of "hallucination" in large models remains a significant issue, making it difficult to achieve definitive results in certain applications [3] - The future of AI in e-commerce will likely involve deeper multi-modal interactions, cross-scenario data integration, and the establishment of new industry norms regarding AI recommendations and user privacy [3]