Xiaohongshu Starts Exploring Large Language Model

According to a May 6 report by 36Kr, Xiaohongshu, a Chinese lifestyle sharing platform, has assembled a dedicated team for large model development since March. The team comprises core employees from the natural language processing technical team of advertising businesses. Currently, the details of this new team are being kept confidential internally.

Zhang Debing, who leads Xiaohongshu’s large language model team, has extensive experience in visual algorithms. He previously served as the head of Xiaohongshu’s intelligent multimedia algorithm team for a year, focusing on AI and audio-video algorithms. Prior to that, he was the head of a multimodal intelligence creation group at Kuaishou, where he oversaw the development of vision-related algorithms.

Along with establishing the large model team, Xiaohongshu has several independent departments dedicated to exploring AI-generated content (AIGC). In April 2023, the company launched an AI creation app called “Trik,” which centers around AI-powered painting. The app utilizes an AI algorithm that generates creative artistic pictures automatically after users upload selfies and landscape photos.

As a content company, Xiaohongshu’s interest in AI models is unsurprising. In the age of PCG and UCG, most internet content companies rely on AI algorithms to distribute information. Xiaohongshu’s content production primarily consists of pictures and videos distributed via algorithms, with commercialization and e-commerce forming the core of their revenue model.

In the era of AIGC, large models based on deep learning can significantly improve content production speed and quantity, subverting traditional content production and distribution methods. With the advent of big models, many Xiaohongshu users have posted about generating content with ChatGPT.

SEE ALSO: Life-sharing Platform Xiaohongshu to Launch Group-buying Feature

Xiaohongshu has a compelling reason to invest in large models, with a diverse and rich community containing hundreds of millions of pictures, notes, user comments, and short videos. This creates a multi-modal database and application scenario that’s ideal for natural language processing and AI-assisted content generation.