Author: Li Donger
The development of large models is in full swing, and many companies have released their own large models. Yunzhisheng, which focuses on AI and intelligent voice, also released a general large model product - Shanhai Big Model on May 24.
It is not surprising that Yunzhisheng launched its own large-model products, and there has been news that Yunzhisheng will release a medical version of ChatGPT. However, compared with the volume of the large model launched by various companies 2 months ago, the large model of Yunzhisheng does not seem to cause much noise.
Launch of Shanhai model in 5 months But customer use is a problem
According to the "Chinese Intelligent Large Model Map Research Report" released by the Ministry of Science and Technology on May 28, at least 79 large models with a scale of more than 1 billion parameters have been published in China. With the release of a large number of large models, people's focus is no longer on whether large models can answer a few questions or calculate mathematical problems, but on which fields these large models can be used, whether they currently have the ability to use them, and whether they can be combined with their own business.
The current application industries of Yunzhisheng Shanhai model mainly include medical, sales, knowledge management, oral language, Internet of Things and other fields.
The medical field is the main service area of Yunzhisheng, and it has previously attracted public "anti-counterfeiting" by industry giants due to its market share in the medical field. However, Yunzhisheng has indeed studied in the vertical field of medical care for many years and has a certain professional data base. At a time when the release of large models is late, giving full play to its advantages in the medical industry to make vertical large models is already a strategy that these latecomer small manufacturers such as Yunzhisheng have to adopt.
Similarly, Yunzhisheng has also accumulated corresponding systems in the fields of sales and knowledge management, and on this basis, upgrading with large models can also provide better services.
In addition to combining existing businesses with AI, the content currently released by Yunzhisheng is mainly focused on NLP (natural language processing). For example, in the Shanhai model, users can use AI to discover and correct their English pronunciation, and conduct conversations to improve their English skills. If it is applied on a large scale, it can help many people who need English listening and speaking practice to reduce a lot of learning costs, and can be widely used in the field of education. In the field of the Internet of Things, Yunzhisheng can make smart home products no longer simply execute commands, but can better understand the natural language spoken by people and interact more closely with human language.
Only from the content released, the Shanhai big model has the basic conditions that a mature large model should have, rich in functions and can meet the basic needs, the accuracy rate is also above the passing line, and many functions have the basis for large-scale commercial application. From the perspective of release time, although the Shanhai big model was released relatively late, it was only later than the industry giants.
According to IDC data, in the first half of 2022, in the Chinese intelligent voice semantic market, iFLYTEK, Alibaba Cloud and Baidu Intelligent Cloud accounted for a total of 25.6% of the market share, Spichi, XiaoI Robot, and Tolls ranked second, but the 3 together can only be comparable to one of the top three in the market share, the specific market share of other manufacturers IDC has not been announced, but the market share of Yunzhisheng included in "Other" is obviously ranked after the above 6.
Source: IDC
However, a problem in front of Yunzhisheng is that it is not the first release of a large model to usher in a turnaround, whether there are enough customers to use is a problem, after all, if you only talk about functions, the capabilities that Yunzhisheng mountain and sea models can provide, other friends seem to be able to provide, even in the medical field, there are many competitors.
Two years without new financing, Yunzhisheng or will be listed again?
AI is a field that requires a lot of time and capital support, which is why most of the leading technology companies are the first to release big models. Yunzhisheng has been attracting much attention since its establishment, except for the fact that no financing news was announced in 2015, until the official submission of the prospectus in 2020, Yunzhisheng has new financing news every year.
But Yunzhisheng's IPO plans did not go as smoothly as its financing. Although the establishment of the science and technology innovation board has given many enterprise technology companies the opportunity to go public and raise funds, the IPOs of many AI unicorns have not been smooth, and some AI companies have later had to temporarily suspend their listing plans, including Yunzhisheng.
However, there may be other reasons for the termination of Yunzhisheng's IPO, among which the public doubts from iFLYTEK are regarded by many as the reason why Yunzhisheng abandoned the listing plan.
The grudge between the two companies may have been buried long ago. Founded in 2012, Yunzhisheng has received a lot of attention because of its founder, but what really made Yunzhisheng the focus was probably the "Hammer ROM" conference in 2013. Since the hammer uses Yunzhisheng's solution, "Yunzhisheng defeated iFLYTEK to become the preferred solution for hammer ROM voice" once appeared in some media reports on Yunzhisheng.
Although it was released for the first time, with Luo Yonghao's Internet celebrity effect, Yunzhisheng quickly gained the attention of LeTV and Sogou and successfully reached a cooperation.
Perhaps because iFLYTEK did not take this marketing slogan of Yunzhisheng seriously, coupled with the fact that it really made a good start, Yunzhisheng was even bolder when it submitted its prospectus in 2020, claiming that it had a market share of 70% in the white electricity market and voice medical record system, and a market share of 30% in the case quality control system.
As soon as this data came out, iFLYTEK's investors couldn't sit still. Since these two areas are also important businesses of iFLYTEK, if according to Yunzhisheng, iFLYTEK has no leading edge in related fields, so investors ask iFLYTEK for relevant data on the interactive platform.
iFLYTEK then responded to investors with data saying that in order to cover the number of hospitals, Yunzhisheng had 10, 36, 91 and 112 in the four reporting periods from 2017 to the first half of 2020, and iFLYTEK had 11, 77, 264 and 489 respectively in the same period. In terms of income, the income of intelligent voice medical records during the above period was 1.7096 million yuan, 9.2639 million yuan, 16.2891 million yuan and 8.9548 million yuan, respectively, and iFLYTEK was 6.6428 million yuan, 29.3727 million yuan, 35.5448 million yuan and 35.7111 million yuan respectively in the same period. In terms of the white electricity market, iFLYTEK also refuted with data, and Yunzhisheng did not reach the level of leading iFLYTEK.
Although Yunzhisheng did not respond to this matter, he chose to withdraw the IPO materials and terminate the listing soon afterward, which inevitably made people doubt Yunzhisheng's confidence.
After the termination of the IPO, Yunzhisheng received another financing of nearly 100 million US dollars in June 2021, but so far, Yunzhisheng has not received new financing for almost 2 years.
However, some people have seen a hint of Yunzhisheng going to IPO again from the frequent appearances of Yunzhisheng in the first half of 2023. For example, in March, an article titled "Future Cloud Zhisheng Listing Can Be Expected?" Cloud Zhisheng empowers the accelerated development of intelligent products", it is clearly mentioned that "Yunzhisheng will continue to make efforts in the three directions of perception and analysis, cognitive technology, decision-making and generation, continue to explore the boundaries of intelligent interaction, give AIoT unlimited imagination with technical advantages, open up more market space, and empower enterprises to develop themselves and achieve listing goals". In an interview with Huang Wei, the founder of Yunzhisheng, Huang Wei said that the valuation of Yunzhisheng should be higher.
Can Yunzhisheng solve the problem of AI enterprise profitability?
However, the main reason for Yunzhisheng's termination of the IPO that year may still be difficult to make a profit. The problem of AI companies not making profits has been criticized.
If the data of IDC is the basis, the top three enterprises are all larger and more mature enterprises, among which iFLYTEK has long been the leader in intelligent voice in China, and the main business of Alibaba Cloud and Baidu Intelligent Cloud does not rely on intelligent voice, and they have all achieved profitability. Although Tolls, ranked sixth, is not as large as iFLYTEK, its main source of revenue is big data software products and services, and it also has the ability to support the research and development of large models.
In contrast, the remaining two second-tier intelligent voice companies, IPO was rejected due to profit problems, although XiaoI Robot was successfully listed, but the profit situation is also unstable, and the stock price has basically been below the issue price since listing.
Although Yunzhisheng has not released financial data after the termination of the IPO, in the prospectus, Yunzhisheng has clearly stated that it will not be profitable in the short term. From the formation of the team to the release of the large model, Yunzhisheng only took 5 months, which inevitably required a lot of money. If new financing cannot be obtained, Yunzhisheng's performance will continue to be under pressure.
Yunzhisheng also has some relatively unique AI technologies, such as sound cloning. The sound cloning of Yunzhisheng is to replicate an AI voice that is almost identical to someone's, and the user can upload the text themselves and generate audio that is read through this voice.
The fire of AI Sun Yanzi some time ago once again aroused people's attention to AIGC in terms of voice. Although Yunzhisheng has not done much publicity about the sound cloning function, there are still many people who want to try it. According to the author's test on the mini program launched by Yunzhisheng WeChat, it took less than half an hour from sound collection to generation, and then entering the text after the generation was completed, and you can indeed get a sound similar to the author's reading.
Perhaps because this function is only used for experience on WeChat Mini Programs, although users can customize voice content, text-generated speech cannot be downloaded, and can only share audio generated by randomly generated text by Mini Program, which has attracted attention, but how to achieve commercialization is a bit puzzling.
Looking back at AI Sun Yanzi's "out of the circle", it is not difficult to find that this is just a reproduction in the music field after the AI painting fire in 2022, although there are some thresholds for operation, but it is not difficult to master. However, Yunzhisheng has no operational experience on the C side, and it seems that it has no intention of entering the C side, and on the B side, the application space of sound cloning in medical treatment, sales and other fields has not been formed. Although this function does have certain application prospects in entertainment and other fields, from the public information, Yunzhisheng seems to have no plans to expand the scope of services to the entertainment industry.
As an AI unicorn, Yunzhisheng must have its own uniqueness, but it was released late and did not intend to commercialize the characteristic business, in the industry of intelligent voice, which is full of giants, it is a little difficult to rely on large models to kill the siege.
(This article is for reference only, does not constitute investment advice, and operates at your own risk)