AI in Action: Which Products Are Genuinely in Use?
Two years ago, the launch of ChatGPT ignited a wave of interest in large-scale AI model development worldwide. Companies across various sectors in China, from established giants to innovative startups, rushed to stake their claims in the burgeoning world of Artificial Intelligence Generated Content (AIGC). As the dust has settled, we now find a plethora of AI products integrated across different industries, making significant strides in user engagement and adoption.
Recent statistics reveal that China boasts around 120 million monthly active users on AI native applications, reflecting a burgeoning user base that has embraced the technology, reaping substantial benefits. Yet, as this two-year period of intense AI evolution unfolds, a pressing question arises: Which products have truly woven themselves into the fabric of everyday life for users? Is the engagement driven by genuine needs, or are we witnessing the rise of fleeting trends?
In discussions with numerous users, it became abundantly clear that in the workplace, the most common applications of AI involve research assistance, document editing, translation, formal writing, and generating images or videos. Industries such as content creation, intelligent customer service, and marketing, particularly those with standardized operating procedures, are seeing heavy utilization of AI tools. Specific products like Wenxiao Yan (formerly known as Wenxin Yiyan) and Kimi dominate conversations in the consumer sector, while Doubao and Tongyi are frequently mentioned by entrepreneurs in the B2B space.
Advertisement
Despite the multitude of AI products available today, a concerning trend has been the lack of innovation in functionality. Issues related to user engagement, retention, and commercialization remain unresolved. While AIGC has been positioned as the next evolutionary step for mobile internet, the anticipated explosion of applications has yet to materialize. Industry experts emphasize that identifying and developing a "killer application" will be crucial for companies looking to prevail in this competitive arena.
The 'Savior' for Workers
Consider the case of Lulu, a corporate employee whose primary responsibility revolves around document editing. She reportedly relies on Doubao from ByteDance and Yuanbao from Tencent, initially using AI to streamline her work through various formal documents. Her aim was to elevate her efficiency, ultimately freeing up more time for projects that enhance her personal skill set.
Lulu expressed satisfaction with her experiences, stating, “Often, I feel a sentence does not flow well, but I’m not entirely sure. It’s far more efficient to extract that sentence and let the AI assist in proofreading it rather than struggling to come up with alternatives or searching the internet.”
Over time, she has noted distinct differences in the AI tools: for writing, she finds Yuanbao to be more user-friendly and relatable, while Doubao suits her needs better for graphic design, offering more options in terms of dimensions and styles.
Another user, Zhang Qing, who operates within the publishing industry, has found considerable utility in AI products for translations and organizational tasks. She comments on how tools like Doubao, Kimi, and Tongyi have significantly eased her workflow.
In her assessment, these products demonstrate intelligence in handling translation, organization, and search tasks. For instance, when translating short to medium-length texts from foreign languages into Chinese, the results are often accurate and fluent, tailored to local expression. However, she pointed out a considerable downside in long texts, where machine translation becomes apparent and lacks cultural nuance.
Zhang elaborates with an example: When collaborating with foreigners via email, Chinese professionals might inquire politely, “What do you think? Can we discuss this in detail?” However, the AI translates this to, “What’s your opinion? Can we have a detailed talk?” Such translations could come off as demanding in an English-speaking context, causing misunderstandings.
Another perspective comes from Zhang Rui, who identifies Wenxiao Yan as her ‘savior’. Her job responsibilities often require her to balance technical work with writing various proposals, summaries, and reflections, demanding a significant amount of her time. For numerous events, such as fire drills or safety lectures, her institution expects comprehensive planning documents and promotional material post-event. During peak times, she finds herself overwhelmed with writing tasks almost daily.
She appreciates that with AI, a set theme allows the system to compile relevant materials logically, complete with citations. For less demanding reports and summaries, she can utilize AI-generated content directly. More intricate tasks can be adjusted based on AI outputs, which saves her significant time. Often, she can finalize a piece within half an hour.
Despite the efficiency improvements brought by AI, there are underlying concerns about privacy. Lulu articulated her worries about data leakage during her interactions with AI, as providing context about her identity and project needs could inadvertently reveal sensitive information. “I’m concerned that my colleagues or superiors might uncover that I used AI for certain tasks,” she noted.
Entrepreneurs’ Goldmine
On another front, while AI tools have streamlined tasks for employees, they have also unveiled lucrative opportunities for entrepreneurs. Zhang Ge Yuan, an entrepreneur specializing in AI solutions, elaborated on how his primary focus lies in B2B operations, particularly in creating AI digital marketing and automation solutions for corporate clients.
Zhang has several successful applications, with AI-driven digital persons being a flagship offering. “We assist clients in utilizing digital personas to generate vast amounts of content and even enable ‘digital person’ technologies to assist in translating short films for international audiences,” he explained.
He highlighted the evolving nature of AI-driven digital marketing, particularly in high-demand tourism sectors, where engaging photos produced by AI serve marketing purposes. Furthermore, his AI automation solutions enable enterprises to house their data securely, while streamlining processes that previously relied on manual efforts like documentation and content review.
Zhang relies on ByteDance’s Doubao model and Coze, an all-in-one AI development platform, to develop his products. This platform's rich ecosystem of plugins and workflows allows users to create bots easily, which in turn form the foundation for minimum viable product (MVP) offerings distributed through various channels.
Moreover, the AI capabilities Zhang employs cover video content with digital personas, voice recognition, and text-to-speech transformations to facilitate diverse content creation. Platforms like Jianying and Jiemeng provide the necessary technological groundwork to fulfill these services, enabling his company to rapidly generate high-quality outputs.
Nevertheless, Zhang warns of the dangerous potential of AI-generated content. He cites the “Three Sheep Recording Incident” as a case study, where the misuse of AI audio-cloning tools resulted in the dissemination of fabricated audio and video content.
A professional approach is essential to mitigate these risks, as the capability of AI to produce indistinguishable content from reality necessitates a cautious strategy moving forward.
Challenges for AI Products and Future Directions
The AI landscape in China is distinctly characterized by various offerings from major corporate players such as Baidu’s Wenxiao Yan, ByteDance’s Doubao, Alibaba’s Tongyi, and Tencent’s Yuanbao, along with promising startups referred to as the “AI Six Little Dragons” like Kimi and MiniMax. Engaging with different users has highlighted the popularity of several of these products, particularly with new data from the “AI Product Rankings,” showcasing significant user interest.
As of September 2024, AI applications have garnered 120 million monthly active users in China. The leading AI products showcasing the highest user engagement include Doubao, Wenxiao Yan, Kimi, Tiangong AI, and Xunfei Xinghuo.
Each product holds unique strengths influenced directly by their parent companies’ backgrounds, as they leverage data accumulated over time to enhance their AI capabilities. For example, Wenxiao Yan performs exceptionally well in document-related tasks through the rich data resources provided by Baidu, while Tongyi effectively utilizes Alibaba’s vast e-commerce dataset for marketing and operational insights.
Looking ahead, industries such as content creation, smart customer service, and marketing are among those predicted to employ AI extensively, complemented by standardized operating procedures optimized by automated solutions. Zhang noted instances in agriculture, where AI facilitates intelligent sorting and pest detection for fruits and blooms.
The release of Qualcomm’s flagship Snapdragon chips in October signals a wave of new smartphones embedded with AI capabilities from brands like Xiaomi, Vivo, OPPO, and Honor. Industry predictions suggest that AI will become a pivotal selling point, enabling seamless connections across devices through models like Xiaoi. Enhanced AI functionalities might even extend into facilitating travel arrangements, such as generating itineraries based on users’ discussions.
Renowned AI practitioner Lian Shilu argues that while the prospects for large models remain promising, the challenges surrounding customer acquisition, retention, and successful commercialization resist resolution. He notes statistical trends indicating low retention rates for several AI applications post-download, with users averaging only around 4.3 days of monthly interaction.
This ongoing narrative has resulted in increased competition that has led to price wars among major AI firms striving to remain valuable in the market. Many have significantly slashed service prices, reflecting over 50% reductions. Such dynamics have prompted a reevaluation of profitability, with gross margins dropping to concerning levels.
Proponents of the “AI Six Little Dragons” suggest these entities face inherent limitations regarding data and computational capacity, making broad expansion challenging. Nonetheless, while the current pace of product innovation may have slowed, the variety of applications and growing open atmosphere surrounding AI leaves room for optimism.
Ultimately, experts conclude that the journey of AIGC is just beginning. The early phase has seen foundational work focused on data curation and model training being accomplished. However, the next phase will shift to unearthing diverse application scenarios. Lian Shilu predicts that the landmark applications of AIGC will materialize by 2026, beyond mere functions confined to mobile devices or smart homes.
Leave a comment