▲Click to jump and like
Recently, the usage of DeepSeek has surged, becoming the fastest application to break through 30 million daily active users.
According to the Ministry of Industry and Information Technology, three major telecommunications companies have fully integrated the domestic open-source large model DeepSeek. In addition to the cloud platforms of the operators, terminal manufacturers such as mobile phones and PCs are also actively embracing DeepSeek, and some local governments have begun deploying DeepSeek in their administrative systems.
What can you do after integrating DeepSeek?
What advantages does DeepSeek offer?
How can we apply the large model
to fully leverage the power of AI?
Major telecommunications operators fully integrate DeepSeek
China Telecom
China Telecom’s Tianyi Cloud is the first operator-level cloud platform in the country to realize the localized inference of DeepSeek. After integrating DeepSeek into the video conferencing system, it can synchronize meeting content, including the voices of participants and the PPT presentations, to form meeting minutes and summarize them into a clear outline.
Lin Binghuai, Chief R&D Expert at Tianyi Cloud Technology Co., Ltd., stated that under the explosive situation of DeepSeek, Tianyi Cloud has unique advantages. The computing power of Tianyi Cloud is distributed across the country, allowing users to access it nearby, achieving very low latency responses.
China Mobile
China Mobile’s cloud platform has customized a dedicated computing solution for the popular DeepSeek-R1 model, providing cost-effective computing support and development environments for innovation in internet enterprises and scientific research in universities.
Sun Shaoling, Deputy General Manager of China Mobile Cloud Capability Center, stated that the DeepSeek model can be automatically deployed on the computing foundation of China Mobile Cloud Platform through the “Computing Network Brain”, reducing the costs of construction, operation, and maintenance of the China Mobile Cloud Platform, ultimately benefiting users by lowering service costs.
China Unicom
On China Unicom’s cloud platform, the programming assistant can quickly generate dozens of lines of code with just one Chinese command, and the DeepSeek model can also provide detailed explanations and optimizations for the generated code.
Experts point out that by integrating DeepSeek, telecommunications operators will fully leverage their strong network infrastructure and massive user data to deeply explore the potential of the DeepSeek large model, providing users with richer intelligent application scenarios and promoting the intelligent upgrade of the entire communication industry.
The circle of friends continues to “expand” as terminal manufacturers accelerate the adaptation of DeepSeek
Technicians introduced that the first Android phone equipped with the DeepSeek model not only significantly improves the efficiency of AI task processing but also brings users a more intelligent interactive experience with deep thinking and personalized services after integrating DeepSeek.
Honor AI expert Bai Chen introduced that after integrating DeepSeek, the accuracy rate of AI answers has improved by an average of 15% compared to before. In some specific cases, the improvement is even greater, with an accuracy rate increase of 85%.
Industry insiders stated that the open-source model of DeepSeek significantly lowers the barriers for artificial intelligence to enter various industries, and related government and enterprise-level applications are expected to experience exponential growth.
Lenovo’s Baiying General Manager Wei Dong stated that at the beginning of the year, Lenovo Baiying and DeepSeek deeply cooperated to first develop an all-in-one machine based on DeepSeek and other technologies, with communication ongoing in government, manufacturing, medical, and education sectors, as everyone hopes to make some local deployments.
It is understood that the Government Service and Data Management Bureau of Longgang District, Shenzhen, has deployed the full-size Deepseek-R1 model on the government extranet, becoming the first government department in Guangdong to deploy this model in the government’s innovation environment. Currently, the capabilities of the DeepSeek model have been opened to various departments in Longgang District.
DeepSeek promotes the transformation and upgrading of chips and intelligent computing centers
The reason why DeepSeek is favored in the industry is mainly due to its innovative algorithms and high cost performance. It not only optimizes the use of computing resources but also brings new opportunities for the development of domestic chips, promoting the industry towards efficiency and sustainability.
Experts point out that DeepSeek’s biggest advantage is its high cost performance. As a super intelligent “brain”, it is not only smart but also very economical to use.
Analyst Bai Runxuan from the CCID Consulting Artificial Intelligence and Big Data Research Center stated that the reduction in the cost of large models has a significant impact on the entire industry, and the demand for chips will change. DeepSeek’s efficient algorithms reduce the reliance on high-end GPUs (graphics processing units), and traditional chip manufacturers like Nvidia may face some market pressure. However, at the same time, domestic chips will encounter new opportunities in customized scenarios.
DeepSeek has significantly lowered the cost of training large models, and training large models no longer requires so many high-end chips and data centers. Will related industries face certain challenges?
Experts indicate that in the short term, some cloud service providers may experience surplus computing power due to model lightweighting, but in the long run, the popularization of AI applications will drive intelligent computing centers to upgrade towards “efficient + green” directions.
In Bai Runxuan’s view, in the future, cloud service providers will push more computing power to edge nodes to support low-latency real-time AI applications, such as autonomous driving and industrial robots. Service models will also innovate, shifting from “computing power leasing” to “AI capability as a service.” For example, operators may provide value-added services such as pre-trained model fine-tuning and industry knowledge base integration.
How to Leverage the Power of Large Models: Professionals Share Tips
In the era of large models, AI technology has moved from the laboratory into daily life, becoming a powerful assistant in work and life. How can we apply large models to fully leverage the power of AI? Professionals share tips:
First, ask precise questions to achieve efficient interaction. Communicating with large models involves technical skills in questioning. For example, if you want the large model to help you plan a trip, don’t just say, “Give me a travel guide,” but provide specific details like, “I plan to go to Changsha for 5 days next month, with a budget of 5000 yuan, and I like food and natural scenery, help me plan my itinerary.”
Secondly, keep optimizing for the large model. Take translation as an example; the initial translation of a large model may lack fluency. Users can provide feedback to further optimize the answers from the large model. For instance, you can point out, “This sentence is not very natural; can you change it to a smoother expression?”
Thirdly, cleverly using prompts is important. Prompts are the “magic words” for communicating with large models, and different prompts can lead to completely different outputs from the large model. If you want to write a science fiction novel, you can use prompts like “interstellar war,” “outer space,” and “aliens” to write a beginning, and with detailed prompts, you can produce more exciting works.
Source: Guangdong Law Popularization Comprehensive Organization, main content sourced from CCTV News
Will you use AI tools like DeepSeek in your life?
Recommended Reading
