On February 21, a screenshot of an internal Emotibot letter was leaked, which mentioned "......Since 2023, business needs have decreased significantly, which has brought serious cash flow pressures and challenges to the company", and announced in the letter that some positions would be suspended for up to 6 months. After a series of spread and fermentation, the news of Emotibot's (comprehensive) shutdown is very loud.
The next day, on the morning of the 22nd, Emotibot issued an official statement through the WeChat public account to clarify the misinterpretation that was circulating. The statement emphasized that the company's current operation is normal, emphasizing that a small number of positions under "partial loss-making business" are involved in the suspension of work and production, and said that it reserves the right to pursue rumors.
According to Emotibot's official website, the products provided by the company are divided into five categories: conversational AI platform, solution platform, knowledge engineering platform, machine learning platform, and natural language processing platform. There are also multiple subdivided products under each category, such as VCA intelligent customer service, multi-modal emotional digital human, Emotibot and other services under the conversational AI platform. At present, there is no information on the specific business lines that are facing the shutdown and reorganization.
However, it can be noted that the statement mentions the current direction of Emotibot's focus, deliberately emphasizing the competitiveness of its new technologies and new products in the AIGC field, reflecting Emotibot's confidence in "keeping pace with the times". Before the outbreak of ChatGPT, Emotibot has been focusing on technologies such as NLP, deep learning, and multimodal emotion recognition, and has achieved the position of unicorn in enterprises in related fields.
It is worth mentioning that at the end of last year's D+ round of financing, Emotibot has announced the official launch of a mature generative AI (AIGC) product using "ChatGPT-like technology", and is about to start a new round of financing.
Of course, the pace of generative AI development is too fast, and even Emotibot, which has been accumulating in the NLP field for many years, seems to be trying to catch up on this new track.
In 2014, Microsoft's R&D teams in Beijing, Suzhou and Tokyo took the lead in launching "Microsoft Xiaoice" in the Chinese market. This "eighteen-year-old artificial intelligence girl" has cooperated with many Internet head platforms, which has aroused widespread attention to intelligent assistants.
As a former vice president of Microsoft (Asia) Internet Engineering Academy, Kan led the technology development of Microsoft Xiaoice. Xiaoice's success made him realize the infinite opportunities of artificial intelligence, and hoped to build "robots that understand the emotions, emotions and intentions of interlocutors like humans" with "emotional intelligence" as the core.
After a year of closed development, Emotibot started the commercialization process and gradually explored a business model based on to B. In 2018, Emotibot was selected into the first batch of innovative artificial intelligence products in Shanghai, which can be said to be the first star enterprise to enter the AI market.
After AI entered the era of large models, Emotibot has also been seeking breakthroughs. In July 2023, Emotibot also reached an AIGC strategic cooperation with Chinasoft International to create a "990,000 Laboratory" to create a large model of industry and scenarios for enterprises. However, as early as in the interview, Jian Renxian talked about the challenges faced by the company in entering the large language model technology from the perspective of Emotibot's business, including the high cost of computing power training and reasoning, high data requirements, high training requirements, high talent density requirements, and the inability to solve enterprise problems.
This may be some clues to understanding Emotibot's current "shutdown" dilemma.
Emotibot's problems are not unique. When the era of generative AI technology represented by large models has truly gained the attention of the whole people, it is not uncommon for the last wave of AI companies focusing on subdivided application fields to face the impact of a new round of AI revolution.
Around 2015, a number of startups with natural language processing and speech recognition technology as the core emerged. Many companies with intelligent Q&A and personal voice assistants as their main services have hastily lost news after a round of financing before they have even revealed their investment.
Even the small I robot, which had already been financed in the D round that year, once fell to 0900 million yuan, facing serious growth bottlenecks, finally went to the United States to complete the IPO last year after a difficult transformation. Founded in 2012, Yunzhisheng, an intelligent voice artificial intelligence company, also suffered a net loss for three consecutive years after 2020, and its Shanhai model was also criticized for its R&D investment far below the industry average. Spire, which has been financing for 15 years to Series B, even paid attention to conversational artificial intelligence in 2007, but is currently being overtaken by many latecomers such as iFLYTEK.
Why do these companies have a first-mover advantage despite having a deep accumulation of NLP technology?
On the one hand, the business of these enterprises is mainly based on B-end oriented intelligent customer service, and the profitability of their products is very dependent on head customers, and their own hematopoietic ability is unstable; On the other hand, when exploring new technologies, it may be difficult for the first mover to adjust the technical route in time, abandon the traditional NLP technology, and may be limited by more customers and scenarios served in the technological innovation, making it difficult for them to make a big impact.
Not only NLP companies, but also the "AI Tigers", led by CV computer vision technology, have also experienced consecutive losses. According to SenseTime's 2023 interim financial report, after the revenue growth rate finally turned negative, the net loss still reached 314.3 billion, the realization of technology after a large amount of R&D investment is still far away. Megvii and YITU, which have not yet completed their IPOs, may face a more difficult financial situation.
In order to keep up with the pace of the times, large models have become something that has to be invested in. But entering the game means that more funds need to be burned, and a lot of manpower and computing resources need to be invested to bet on an uncertain future. Although the top players in the AIGC field are far from reaching the stage of profitability, even Altman himself has expressed his hope on X to raise a staggering 8 trillion yuan to invest in the development of AI chips. However, the problems faced by established AI companies seem to be more serious, and if they can't shake off the burden of history and fully embrace new technologies, perhaps the embarrassment of adapting to the soil and water in the era of large models will always be with us.
The giants of the next era are often not the leaders of the previous era. "In the world of technology, it's not just an inference of laws, it's like a curse.