Based on details provided in the announcement, the platform is set to integrate past industrial practice and technology accumulation, with the company having announced ChatJD's landing application roadmap "125" plan, with the power of industrial AI, to accelerate AI technology implementation and development in China, and promote the development of the real economy.
A milestone within AI technology development, the GPT pre-training model helps optimise the technical route of natural language and is seen as an innovative paradigm within the field of intelligent communication. Leveraging algorithms and RLHF (Reinforcement Learning from Human Preferences), ChatGPT has high fluency, reaches a high conversational level, and can realise text generation, code generation, document translation, and business copywriting, in addition to a series of common output tasks.
As a bottom-layer dialogue system, the ChatGPT technology demonstrated strong capability when talking about versatility, but it is still lacking in terms of loyalty, credibility, and accuracy, predominantly due to the lack of knowledge in the vertical industry and middle-layer field, as knowledge presents difficulties for wide implementation in the real application layer.
Following the needs of the industry, JD Cloud's Yanxi artificial intelligence application platform will launch ChatJD, looking to create an industrial version of general ChatGPT with advantages, high frequency, and rigid needs.
As detailed in the announcement, ChatJD is believed to quickly reach the standards for landing applications via deep cultivation in vertical industries and is also set to continue promoting the generalisation between different industries to create increasingly general-purpose industrial versions of ChatGPT. Furthermore, the aim is to build a ‘flywheel’ of data and models to subdivide real and professional scenarios, improve the platform’s capabilities, and offer feedback and refine the industrial application capabilities of the general ChatGPT.
one platform: ChatJD intelligent human-computer dialogue platform for understanding and generating tasks in natural language processing, having an estimate of hundreds of billions in parameter volume;
two fields: retail and finance. Due to JD Cloud’s accumulation of real scenarios in the retail and financial fields for over 10 years, it has a four-layer knowledge system, over 40 independent systems, more than 3,000 intents, 30 million high-level Q&A knowledge points, and an ecommerce knowledge map that covers over 10 million self-operated products, increasingly vertical and focused;
five applications: content generation, man-machine dialogue, user intent understanding, information extraction, and emotion classification, covering the most re-used application scenarios in the two industries, as well as customer consultation and service, marketing copy generation, product summary generation, and ecommerce. Fields of the likes of live broadcast, digital human, research report generation, and financial analysis are believed to play an extensive range of landing values.
Every day we send out a free e-mail with the most important headlines of the last 24 hours.
Subscribe now