The intersection of artificial intelligence and social media platforms has given rise to a fascinating phenomenon, especially in the context of Xiaohongshu, a popular Chinese lifestyle platform. Recently, a noticeable surge in AI-generated content has caught the attention of users, leading to a mixed bag of reactions ranging from awe to skepticism.

Many of these accounts use catchy monikers like "Virtual Girlfriends" or "Cyber Angels," and despite the platform marking some images as "suspected AI-generated content," users actively engage with them, showering these accounts with compliments and requesting fashion tips even when they know the images may not depict real people.

The technology behind generating these AI images has matured significantly, achieving a realism that blurs the line between reality and illusion. The painstaking preservation of minute imperfections has allowed these digital creations to resonate with human emotions, perhaps explaining the rampant popularity of AI-beautified accounts on Xiaohongshu.

This phenomenon has birthed a unique "AI debunking niche," where users take it upon themselves to identify those AI accounts that have slipped through the platform's verification processes, falsely presenting as genuine influencers without revealing their digital origins. Techniques for distinguishing humans from AI images have become popularized—observing "eye reflections" in photos being one amusing example.

Advertisement

In addition to stunning AI-generated personas, the platform has seen a rise in the "AI Grandmas," who use tools to churn out cheerful nuggets of wisdom and motivational quotes.

These AI Grandmas are often adorned in charming anime-style designs and are programmed to dispense words of wisdom via sleek snippets combined with delightfully whimsical visuals, effectively drawing in crowded audiences. Once they’ve established a solid following, monetization kicks in through sales promotions and advertisements.

While AI-generated beauty is stealing some of the limelight, a different kind of AI content is also flourishing on Xiaohongshu, focusing on educating users about AI technology itself. If one wishes to learn about leveraging AI for writing, resources abound—from lists of “six indispensable free writing tools” to comprehensive tutorials and community support, as well as personal diaries detailing experiences with specific applications.

An impressive spate of AI professionals has begun to utilize Xiaohongshu as a critical platform for sharing cutting-edge insights and establishing a loyal follower base in a relatively short period.

 Reports suggest that while the site recognizes the need for effective management of AI-generated content, there’s also a drive to embrace AI-centered narratives that resonate with targeted audiences.

As Xiaohongshu's Vice President Xu Lei articulated at the 2024 Wuzhen World Internet Conference, AIGC (AI-generated content) has drastically lowered the barriers to content creation; however, it concurrently raises a myriad of governance challenges.

Xiaohongshu's approach appears to be a dual focus; having already laid the groundwork for a governance framework, the platform is committed to maintaining "truth" as the lifeblood of its existence.

In reality, Xiaohongshu is quite keen to capitalize on the opportunities stemming from this AI wave.

The platform is spearheading creative content campaigns within its community, launching events focused on AI from October onwards, and articulating, "If you are as excited about AI as we are, consider Xiaohongshu your first stop in this adventure." 

Furthermore, the campaign #AINewbieVillage# invites users to share AI tool recommendations, tutorials, and personal experiences, aptly targeting those eager for insight or entertainment.

In an era dominated by competition in AI large models, Xiaohongshu intends to remain an active player rather than a bystander.

By March 2023, Xiaohongshu had initiated the creation of an independent large model team, with key personnel from its NLP technology arm in advertising. Led by Zhang Debing, a professional with years of experience in AI and multimedia algorithms, the new team aims to integrate multiple modes of technology into Xiaohongshu’s unique framework.

By October that year, Zhang shared insights at the Xiaohongshu REDtech Youth Technology Salon, outlining the plans to pivot the large model strategies towards three key areas: implementing multimodal technologies in internal applications, leveraging AIGC tools to enhance efficiency for content creators, and focusing on cost control.

This year, the platform has also embarked on gray-testing its internally developed universal large model base, dubbed "Little Sweet Potato," across select products.

Following the app's suspension, Xiaohongshu quickly beta-tested an AI illustration feature called "This Moment" and a doodling function. Users can input text or sketches, allowing the AI to generate corresponding cover images. While the "This Moment" module has been removed from the platform, the illustration generation feature remains, albeit with a more entertainment-focused lens.

Nevertheless, the generated images are often simplistic and do not fully meet users' needs when crafting content, leaving room for further enhancement.

The offerings include an emotionally supportive "AI Pet Companion," as well as more entertainment-focused features like "AI Character Generation" and "AI Voice Acting." Additionally, tools like "Da Vinci" and "Captain Potato's Assistant" cater to the organization and interactivity of user-generated content.

The "AI Pet Companion," based on a MiniMax model foundation, is a social engagement tool allowing users to create personalized pet characters that can interact and accompany them, reminiscent of the once-popular QQ Pets concept, albeit with more interactivity.

Users can pet their AI companions, visit other users' pets, and engage in light-hearted conversations, with reminders for daily fortune readings. However, relying on a database predominantly informed by Xiaohongshu's own corpus can lead to relatability, while some aspects may elicit a negative response during interactions due to pronounced emotional tones.

Users have noted instances where their AI pets exhibit unexpectedly negative thoughts or opinions, even abruptly ending conversations when posed with complex inquiries.

Known as Xiaohongshu's equivalent to ChatGPT, "Da Vinci" began internal testing last September and officially launched to the public mid-year. A straightforward query will prompt it to provide structured, detailed responses accompanied by links to "possibly interesting notes."

Similar to the recently tested comment assistant "Captain Potato's Assistant," it excels at summarizing and clarifying information rather than generating creative content and collecting extensive external data. The convenience in light usage contexts, minimizing repetitive searches for note explanations, has made it appealing for users.

When prompted, Da Vinci emphasizes its role in "helping you find the notes you need," thereby enhancing content flow back into Xiaohongshu's ecosystem.

Leveraging Xiaohongshu's extensive content database, Da Vinci showcases remarkable proficiency when addressing popular community inquiries related to fashion tips, travel guides, or health regimens, often providing more tailored suggestions compared to other AI systems.

With many users utilizing Xiaohongshu as an alternative search engine, the AI-driven search functions raise questions about enhancing Xiaohongshu's capabilities. While currently in the testing phase, the "Search Potato" feature has yet to pass the effectiveness test after some user requests resulted in overly engaged and biased viewpoints.

By June 2023, Da Vinci's introductory phase hinted at its enigmatic nature with prompts like, "If you discover me, kindly keep my existence a secret."

Insiders reveal that Xiaohongshu’s method for new AI products is characterized by extreme caution, fearing disruption to the existing content ecosystem, thereby posing the dilemma of reconciling genuine human engagement with AI-generated content.

The algorithmic strength in search, together with stellar original visuals and rich comment threads, generates an expansive corpus of data.

However, the real challenge lies in optimizing this large model: precise direction in optimization is the crux of Xiaohongshu's dilemma. The Da Vinci case highlights this complexity—while raising the level of intelligence could divert traffic away from original content creators, a lack of sophistication could repel potential users.

When Xiaohongshu constructs large models with its unique homegrown content, creators behind original work find themselves filled with concern over intellectual property rights.

In August 2023, Xiaohongshu's AI art application Trik faced accusations of infringement from traditional artists, leading to a legal battle that has since progressed through the courts. One party affirms, "We are not opposed to AI's growth; we merely want to safeguard human creators' fundamental rights."

The ongoing discourse within legal and industry circles surrounds whether using existing works for model training falls within reasonable use parameters.

Ultimately, as the winds of AI begin to sweep across Xiaohongshu, the platform must navigate a challenging path toward optimizing its potential and resolving the inherent complexities along the way.