Article Source: AI Whale Selection
Author: Endge
Image Source: Generated by WuJie AI
In the wake of OpenAI’s Sora becoming extremely popular and the launch of Google’s Gemini, we’ve seen domestic firms specializing in large-scale AI models struggle as they stand on the brink of commercialization.
On February 21st, the well-known AI startup Zhujian Intelligence announced a sudden six-month halt to its operations, citing a significant reduction in business needs that led to intense cash flow pressures and challenges.
Zhujian Intelligence was founded in 2015 by Jian Renxian, former Vice President of the Microsoft (Asia) Internet Engineering Academy, and boasts proprietary technology for human-computer interaction through text, image, and voice, as well as a self-developed “multimodal emotional recognition model.”
I visited their booth at the 2023 World Artificial Intelligence Conference in Shanghai – it was bustling with activity. Yet, just six months later, the company entered a state of “shock.”
Zhujian Intelligence’s Shining Moments
Many AI innovation companies are located in Shanghai’s Caohejing, including ByteDance, SenseTime, Singularity Technologies, and Zhujian Intelligence. Among these, Zhujian Intelligence was a particularly well-developing enterprise.
Looking at their fundraising history, they had raised over 100 million yuan almost every year, totaling over $100 million and 4 billion yuan – an impressive fundraising feat surpassing 1 billion yuan.
Zhujian Intelligence developed a range of NLP technology platforms such as Bot Factory, Knowledge Factory, and Emoti Coach. Based on these platforms, the company launched a series of products that have been adopted by hundreds of clients across banking, securities, insurance, manufacturing, and consumer industries.
Anticipating the arrival of the era of large models, Zhujian Intelligence unveiled its “1+4” product strategy in 2023, powered by AI’s large language models.
Zhujian Intelligence’s 1+4 Product System
This system essentially builds the last mile of commercialization for domestic large models. On their EmotiBrain large model, one can choose from general-purpose foundational models like Huawei Pangu, MOSS, MiniMax, LlaMA2, and Baichuan. They employ fine-tuning techniques to train bespoke models.
Thanks to their NLP technology, Zhujian Intelligence helps businesses better train with proprietary knowledge. With its products like KKBot, AI Coach, Knowledge Factory, and Magic Writer, they find applications in marketing, customer service, sales training, knowledge management, and copywriting scenarios.
This means that in helping businesses apply large models, they can choose better large models to train corporate-specific ones, integrating proprietary knowledge for enhanced performance in various AIGC applications.
From their messaging, it’s apparent that Zhujian Intelligence aims to create a “model factory” to help 90% of businesses deploy AIGC within a year.
A Sudden Collapse Before Dawn of Commercialization
From their public information over the past couple of years, it seemed they still had a substantial number of collaborators.
In March 2023, Zhujian Intelligence announced strategic partnerships with Inspur and Zhipu AI to further promote the application and commercialization of AI technology.
Moreover, they were a deep partner of Huawei Cloud and played a core role in Huawei’s Pangu large model project. The COO of Zhujian Intelligence described it as an excellent partner for the AI industry’s landing with Huawei Cloud.
With top-tier partners like Huawei and Zhipu, coupled with downstream clients including China Minsheng Bank, GAC Group, and China UnionPay among over 300 large enterprises, and over 50 ecosystem partners reaching hundreds of millions of end-users, the company seemed well-positioned.
In the era of large models, CEO Jian Renxian of Zhujian Intelligence emphasized that enterprises don’t need excessively large models. “Our general-purpose large model and model factory deployed to clients, combined with their data and industry know-how for parameter adjustment, allow for fine-tuning to create a 7B-65B scale model.”
The problems the company faced might stem from its origins in NLP; as the ChatGPT wave hit, Zhujian Intelligence’s reluctance to move away from NLP might have caused business difficulties.
Thus, in an internal open letter, the company mentioned realigning its business direction, prioritizing existing customer service, and exploring new technologies and products. They would also optimize loss-making business lines, with certain departments and positions ceasing operations immediately for restructuring.
Large Models and AIGC Aren’t Cure-alls
The enterprise application of large models is still in a wait-and-see phase, and the costly fees for private large models, reaching up to 2-3 million yuan, aren’t something every company can afford on-demand. Although cheaper, Zhujian Intelligence’s model still required corporate recognition and decision-making to catch up.
However, following their last D+ round of funding just before December, the company couldn’t ride the wave to commercial success before announcing a sudden halt in operations.
Zhujian Intelligence left the door open for its employees to return:
Beginning from February 20, 2024, we will halt operations for six months, with the downtime scheduled from 10:00 AM on February 20, 2024, to 7:00 PM on August 19, 2024. Throughout this period, we will continue to monitor market conditions, actively seek opportunities for business development, and ensure care for every team member. The company will keep you informed and provide timely updates regarding returning to work or other related arrangements.
Wage standards during the work suspension
In accordance with Article 12 of the “Regulations issued by the Shanghai Municipal Human Resources and Social Security Bureau”, from the first day of the holiday until the end of one month’s wage payment cycle, employees are to be paid as agreed.
After the holiday exceeds one month, the company shall pay the minimum wage standard of Shanghai.
Currently, only a few companies like Baidu, Alibaba, and heygen (an overseas digital person company) have managed to keep revenue at hundreds of thousands of dollars a month.
Many more companies in the large model and AIGC space are still struggling to find markets where the commercial effects stand out, leading to reputation without profitability.