DeepSeek burst onto the scene and became the undisputed crowd-puller during the Spring Festival. The shock it caused in the industry still resonates. Today, DeepSeek has become the new engine for improving the quality and efficiency of software products. It brings new impetus to the development of industrial software companies and is a new productive force for the high-quality development of the industrial software industry.
DeepSeek: Enables One-Click Group Creation – A Global “Friends Network”
DeepSeek is like a spring breeze that makes the “southern shores” of the industry value chain green. The entire industry is working feverishly to adapt. It seems as if DeepSeek has activated one-click group creation – suddenly it has friends all over the world. International technology groups such as NVIDIA, AMD, Microsoft, Amazon Web Services and Intel have already announced their connection to DeepSeek model services. In China, cloud platforms such as Tencent Cloud, Tianyi Cloud, Alibaba Cloud and Baidu Smart Cloud have officially announced their support for DeepSeek. Chinese GPU companies such as Muxi and TianShu AI are also adapting their systems to DeepSeek.
More than 15 domestic chip manufacturers, including Huawei Ascend, Biren Technology and Kunlunxin, have announced their adaptation or listing of DeepSeek model services. Some industrial software companies have also begun integration. As early as May 2024, Zhongkong Technology used the world’s leading AI model version DeepSeek V2 as the core model for an industrial AI product in the BA field to promote the use of AI in research, production, supply, sales, service and support for enterprises. This product is another “nuclear powerhouse” for Zhongkong Technology’s digital and intelligent industrialization after the release of TPT and is expected to be launched this year.
With the release of the more advanced DeepSeek V3 and R1 models, Zhongkong Technology has quickly integrated and upgraded the basic AI models. By leveraging its excellent inference capabilities and openness, logical reasoning has been greatly improved and operating costs have been drastically reduced. On the basis of DeepSeek, Zhongkong Technology developed the “Industrial BA Super Intelligent System”, which includes five multi-modal perception centers, industrial BA graphics data and neural networks, intelligent industrial app creation, multi-intelligent collaboration for the entire business chain, and an industrial analysis and decision-making center.
DeepSeek advances three fundamental reconstructions
Chu Jian , founder of Zhongkong Group and Ningbo Industrial Internet Research Institute, believes that DeepSeek offers three key opportunities: accelerating intelligent transformation, innovation in software development model, and innovation in business model.
First, DeepSeek brings a tremendous acceleration of intelligent transformation. By integrating DeepSeek’s multimodal and powerful inference capabilities, the boundaries of intelligence in the areas of execution, management and decision-making can be greatly expanded. Industrial software equipped with DeepSeek can significantly reduce operating costs and increase efficiency with digital workforces. DeepSeek’s open source model and low-cost operation also provide a favorable environment for the commercial implementation of intelligent industrial software.
Second, DeepSeek revolutionizes the industrial software development model. Its strong inference and code generation capabilities can profoundly change the research and development model of industrial software companies, improve user experience, and better support business changes.
Third, DeepSeek promotes business model innovation. Industrial software companies can use DeepSeek to optimize intelligent products and services, provide data-driven decision support, build platform services, and improve product lifecycle management to adapt to market changes, increase service value, and ensure the sustainability and competitiveness of their business model.
DeepSeek has a profound impact on the industrial software industry. Chu Jian predicts that traditional industrial software will be replaced by intelligent agent clusters in the near future and three fundamental reconstructions will occur:
- Reconstruction of the development model for industrial software : The traditional rule-based development process is replaced by an AI-native development approach. With DeepSeek’s powerful inference and code generation capabilities, industrial software can be developed more efficiently and cost-effectively.
- Reconstruction of the user experience : Traditional process-oriented systems are replaced by goal-oriented systems. By integrating DeepSeek, industrial software is transformed from manual operation to human-machine multi-agent collaboration and converted into a multi-agent system (MAS).
- Reconstruction of decision control : Traditional rule-based data presentation is replaced by control logic based on user intentions. With DeepSeek’s powerful planning and inference capabilities, the decision-making process is redesigned through human-machine collaboration.
DeepSeek brings with it three major challenges
“In addition to the technological impact, the sudden success of DeepSeek has promoted the spread of AI throughout society and created a broader awareness of artificial intelligence,” explained Chu Jian. “At the same time, it has created new challenges, particularly in three areas: data accuracy, business model transformation and skills shortages.”
First: data precision – the highest accuracy is required. In industrial applications, the requirements for accuracy are extremely high. Systems based on large models such as DeepSeek must meet the required data approval standards in all application scenarios.
Second, the transformation of traditional industrial software companies is inevitable. The change can no longer be stopped – business models, development strategies and delivery processes must be comprehensively restructured. For companies, this means profound, systematic innovation and a redesign of the organizational structure.
Third, the lack of interdisciplinary professionals is a critical factor. There is a lack of talent with both industry expertise and knowledge of data science and AI. Transforming industrial software using large models like DeepSeek requires skilled professionals – from model conception to product development and technical implementation.
In addition to DeepSeek, SUPCON Technology has already introduced a number of other open source models, including ChatGLM-9B, Qwen-13B, Qwen-30B and Qwen-72B. But which of these major models offers the best reasoning capability within a chain-of-thought structure? And which model is the best choice for agent applications?
Chu Jian explained: “Looking at the characteristics of the models, Qwen-72B shows outstanding performance in terms of dialogue stability and logical reasoning within a chain of reasoning. At the same time, we extensively use OpenAI’s ChatGPT-4o and ChatGPT-o1 series. For agent scenarios, ChatGPT delivers excellent results. In general, the larger the model parameters, the better the question-answering capabilities – but this also increases the operating costs. Smaller models such as ChatGLM are acceptable in answering questions, but have a slightly weaker reasoning ability for agent applications.”
Author: Yang Guang Source: China Informationization Weekly