GuShiio.com Wen| Newberry daybreak, author| He Lingzheng, Editor| Zhai Wenting
The “DeepSeek Storm” sweeping the world is still indiscriminate in all walks of life, and the anxiety index of migrant workers has exploded.
He had just returned to work during the Spring Festival holiday, and Li Yang saw several messages about “AI training” in the law firm’s group. There are “New Practices in the Development of Technology-Empowering Lawyers ‘Business-DeepSeek Application Special Lecture” sponsored by the Chengdu Lawyers Association, and there is also a live broadcast invitation for “DeepSeek Legal Person Practice” sponsored by a certain industry website.
Most of the non-traditional employment relationships between lawyers and law firms, so law firms often do not require participation in training, but Li Yang did not want to miss it. Due to time considerations, he chose the latter and entered the live broadcast room at 19:00 that night. During the one-and-a-half-hour live broadcast, the number of viewers remained at around 130,000. Assuming that they are all practicing lawyers, they account for 17% of the country (750,000 practicing lawyers nationwide in September 2024).
Li Yang was a little moved by this number.”To be honest, if it weren’t for the training, I would never have even heard of this website in all my years of practice.” What he knows is that several law firms have recently been promoting DeepSeek training. Li Yang once used DeepSeek to write a defense opinion.
In the past month, the disruptive innovation brought by DeepSeek has also cast migrant workers in the shadow of being replaced by AI. Every industry is thinking about how to improve the “AI rate” in business processes.
Not only do ordinary people feel anxious, but even local managers cannot let go of missing out. Not long ago, Jiangsu official media published three special articles for two consecutive days: “Why did DeepSeek appear in Hangzhou?” “Why couldn’t Nanjing develop the ‘Hangzhou Six Little Dragons’?” “Hangzhou has DeepSeek, what has Nanjing?” A few days ago, Shenzhen’s Futian District launched AI digital employees based on DeepSeek, which sparked heated discussions.
A top-down national AI transformation movement is unfolding.
Overnight, from useless ribs to must-have
Just one Spring Festival, employees of state-owned enterprises in third-tier cities traveled overseas and felt that AI had changed from a “unnecessary” technology to a “must-learn” technology.
In the first week of construction, Transeyang received a group of “DeepSeek Training Notice” from the company. This training was initiated by the group headquarters, and third-party training companies and branches were invited to sign up and participate in the training in turn. The main content includes the popularization of basic AI knowledge, analysis of the characteristics of various AI tools, and how DeepSeek+ various efficiency tools can assist office work.
Unlike lawyers who can attend training selectively, Transoceanic companies require everyone to attend.
For DeepSeek, Transocean has heard about it during the Spring Festival. After trying it, I was a little surprised and shocked. She just didn’t expect that as soon as she returned to work, she would be instantly swallowed by this wave, and her anxiety exploded. Before the Spring Festival, everyone still felt that AI was a little useless.
Chang Jia was also deeply touched by the dark surge of AI.
As a reporter for a newspaper in Chengdu, she recently felt that the leader had become like a “class teacher”, requiring all employees to learn to use DeepSeek and also check the learning results.
Chang Jia was the only person in the newspaper who had previously run the AI line. The worrying profitability of local AI companies failed to bring her more business cooperation. Before the Spring Festival, leaders thought about giving up this line, but the situation turned sharply after the year.
Chang Jia believes that the change in leaders ‘attitude towards AI does not come from AI, but simply because of DeepSeek itself. He even requires that “every reporter publish an actual manuscript related to DeepSeek every week.” Editors at newspapers also began using DeepSeek to write headlines.
Because of his certain research foundation in AI, Chang Jia naturally received an “exemption” from learning. But she has an invisible additional task: teaching everyone to use AI.
Like a row of dominoes being knocked down, learning DeepSeek overnight has become a trend across the country. The first lesson after construction in many places is to train DeepSeek. Some of them are organized by government agencies in Puyang, Wuhan, and Ningbo, and some are hosted by enterprise associations.
AI invades thousands of industries as fast as a virus. Especially in industries where people themselves produce resources, such as real estate brokers.
Liu Jin, an employee of Lianjia, recently received an oral notice from “Shangfeng”: Learn to use DeepSeek. Although there was no formal training, Liu Jin observed that all colleagues in his store took the initiative to embrace DeepSeek. The reason is also very simple: the internal system of Lianjia is not as effective as DeepSeek in searching for listings.
“Most of the requirements put forward by customers are very vague. If we use our system, we may not be able to find hidden information after a long time. Now, as long as we tell DeepSeek information such as budget and needs to, the communities we find are often more suitable.”
Anxiety push
Ordinary people may not be the first to feel the shock wave. After the release of DeepSeek V2 open source model, it became the “price butcher” of large models, quickly triggering a price war in the large model industry. Big manufacturers such as Byte, Alibaba, Baidu, and Tencent have followed suit.
Founder Liang Wenfeng’s idea is,”AI should be universally beneficial and affordable to everyone.”
But just as the source is striving to reduce the cost of AI use, a new revenue generation is spreading invisibly.
As DeepSeek’s popularity soars, corporate anxiety also heats up. From senior executives of large factories to owners of small and medium-sized enterprises, they have begun to think about how to improve the “DeepSeek rate” in business processes, and even try to link it with KPIs.
For white-collar workers in first-tier cities who are familiar with various productivity tools, there may be no threshold for using DeepSeek. Even if you have doubts, you can get more free learning resources through the Internet. “DeepSeek From Beginnings to Proficiency”, which was previously rumored on the Internet, was produced by Tsinghua University.
However, because the whole society is permeated with a “FOMO sentiment”, and some influential business figures have invisibly fueled the flames. Some people began to smell the business opportunities.
Traffic bloggers from all walks of life have made the scene, earning large waves of traffic through technological innovation and national sentiment, and have been popularizing DeepSeek to rural areas.
During this period, various course-selling accounts and training institutions used inflammatory titles, such as “If AI is about to eliminate XXX occupations/industries, you will fall behind if you don’t use it anymore” and “XXX used AI and earned XXX thousand yuan in a month” to win traffic and sell classes at the same time.
Currently, the prices of DeepSeek courses on the market range from tens of yuan to thousands of yuan, and some have sold more than 8000+ per month. But the content is uneven. Some institutions have changed the prompt words of ChatGPT for the past two years to DeepSeek. After completing the program, some even commented that “asking DeepSeek is better than buying this course.”
The value of enterprise-level training is also not satisfactory. At present, most corporate training on the market still focuses on “DeepSeek+ Various Tools”. This kind of content is explained on various video websites, and the training institutions just make an integration. For example, this is the case with the live broadcast class for legal persons mentioned above. Even the class materials given by the other party are directly in the PDF of Tsinghua University.
Fortunately, there is no charge for legal persons to live broadcast. Lawyers who want to participate only need to forward the poster to their circle of friends, attach the sentence “forwarded”, and then feedback the screenshot to the staff.
However, training is only for the beginning, and ultimately for selling goods.
According to Li Yang’s recollections, the entire live broadcast was highly marketable. Every once in a while, the female anchor would jump out to promote the website’s annual member, 3.3 years/2700 yuan. Assuming that 10%(nearly 13,000 people) of the viewers placed orders that night, that would amount to tens of millions of sales.
DeepSeek detonates AI anxiety
Before DeepSeek was born, many AI tools had entered the public eye.
The emergence of ChatGPT-3 two years ago seemed to pull the world out of science fiction. This was followed by two years of the big model arms race. Kimi, Doubao, and Tongyi Qianwen, which have been born one after another, have all competed on the download rankings, but none of them have caused such a huge wave.
A survey report by Ai Media Consulting in 2024 mentioned that nearly 60% of users said they would not directly use the generated content of the AI model and would modify it to varying degrees.
For most work, manual correction is indispensable, and there is no way to replace it. The Paper News showed in the “2024 Public Attitudes Survey Report on Artificial Intelligence” that only 8.0% of people believe that artificial intelligence has more or only negative effects on themselves.
Human beings ‘unique ability to express emotions and reason about complex things was once a “safety barrier” in the hearts of many people. Until the emergence of DeepSeek, the validity period of this barrier was greatly shortened.
The reasoning speed and writing performance of the DeepSeek R1 model are amazing. Coupled with the extreme cost performance, the cost of computing power has become a “floor price”, shaking Silicon Valley. Within 18 days of launch, DeepSeek reached 15 million daily accounts. In comparison, it took 244 days for ChatGPT to reach 15 million daily active lives.
There is no active streaming or deliberate publicity. This is the first time that generative AI has occupied the C-side at such a speed. For comparison, Byte spent 1 billion yuan on the C-terminal last year to promote bean buns, which is 100 million yuan more than Kimi’s estimated total investment amount. But entering February, the daily activities of Doubao and Kimi fell one after another.
While most ordinary people are still cheering for this technological innovation, the big factory can no longer hold back.
On February 1, the fourth day of the Lunar New Year, Huawei Cloud officially announced the launch of DeepSeek-R1/V3 reasoning service based on Huawei Cloud Shengteng Cloud Service before accepting leave. Then there are Alibaba, Tencent, Baidu, and Byte, which access DeepSeek to varying degrees.
Among them, Tencent’s embrace is the most complete. In addition to business-side products such as WeChat, Tencent Docs, and QQ browser, Tencent’s self-developed AI assistant “Tencent Yuanbao” has also been officially connected.
Dachang’s anxiety is evident in words. Taking Byte as an example, although DeepSeek is only connected through the volcano engine, the bean buns have not been moved for a long time. However, Liang Rupo bluntly stated at the staff meeting on February 13 that the company was “not sensitive enough to major changes in AI technology” and admitted that it was lagging behind DeepSeek in the follow-up speed of the long-chain thinking model.
Dachang’s attitude was also like a huge rock, crashing into the lake where there were only a few small water sprays. In less than a month, more than 200 companies connected to DeepSeek. The rapid response at the enterprise level is gradually transmitted to individuals like ripples, and the threshold for the use of AI quickly collapses.
“Large models assist in writing papers”,”Agents write copywriting in batches”, and “DeepSeek assists in prescribing traditional Chinese medicine”, more and more use methods are developed by individuals. On Xiaohongshu, more and more “AI aborigines” born after 2010 appear to teach everyone to use AI.
At the same time, a sense of crisis that had no time to think carefully also spread. Turning on social media, practitioners in many industries are expressing anxiety about being replaced. Including occupations that were once in “safe areas”, such as psychological counselors and traditional Chinese medicine practitioners.
Not long ago, a screenshot of a wildly circulated chat history pushed this sense of crisis to a small climax.
In the screenshot, Shangmei CEO Lu Yixiong said in an internal group that AI would be used to replace most positions. Among them, the customer service department was cut by 95%, leaving only 5%; the content innovation department was cut by 80%; the new product innovation center was cut by 70%; and the legal department was cut by 50%.
Although Lu Yixiong later clarified that “only some places add people and some places reduce people,” the screenshot still caused quite a bit of controversy on social media, especially the sentence “One team has to do the work of the previous 20 teams”, which made netizens sigh that employees who were laid off and who were not laid off have become “DeepSeek victims.”
Shangmei is not the first company to propose layoffs due to AI, and it will not be the last. For operators, layoffs are to reduce costs and increase efficiency, and will also build a new organizational logic.
In the AI era, this may be just the beginning. Everyone must be prepared for the shock wave.
