The 2024 Government Work Report first proposed the "Artificial Intelligence+" strategic action, aiming to promote the empowerment of thousands of industries with artificial intelligence. The new generation of artificial intelligence technology represented by big models has become the core driving force for the intelligent transformation of the software field. Big models inject new intelligence into traditional software, bring new capabilities and changes to software and its forms, and also bring new ideas to enhance software research and development productivity and accelerate innovation, and promote structural changes in the software industry. The value brought by large models to the software field enhances production efficiency and reduces project risks. The application of intelligent development and testing tools for large models provides more automated and efficient methods for software production processes. For example, through the abilities of code generation, code completion, and Q&A, software developers can write high-quality code more quickly, significantly reducing manual debugging and error repair time. At the same time, the use of such tools can reduce the dependence of projects on developers. Based on the learning and memory abilities of the tools, they can not only assist new developers in quickly developing code that meets project requirements and specifications, but also help developers learn new programming languages quickly, thereby reducing project risks caused by personnel turnover. According to CSDN's 2024 survey data, AI technology has become an indispensable part of the work of software developers in China. 69% of developers say they are using AI tools, and 38% of developers believe that AI coding aids can reduce their workload by 20% to 40%. Improve software quality and enhance product stability. Through intelligent code inspection and intelligent unit testing capabilities, developers can quickly perform code validation and testing, timely discover and solve potential problems such as code defects, code exceptions, code security risks, etc., thereby helping developers write higher quality code, reduce software failure rates after release, and improve software stability and performance. GitHub research data shows that Copilot can help developers fix over two-thirds of vulnerabilities during the coding process. In addition, the generation and completion of test cases and test data can improve the coverage of system testing, thereby enhancing the quality of testing and reducing software defects. Large models can serve as powerful assistants for software developers, providing intelligent capabilities such as understanding and generation for various stages of development, helping to reduce costs and increase efficiency. Accelerate software innovation iteration and enhance enterprise competitiveness. By improving development efficiency and enhancing software product quality, enterprises can launch innovative products faster, seize market opportunities, and stand out in fierce market competition. On the one hand, the intelligent R&D process shortens the software production cycle, allowing enterprises to adapt more flexibly to market changes and accelerate product iteration rates; On the other hand, the assistance of intelligent R&D tools enables developers to free themselves from repetitive and tedious low-end coding work, thereby having more time to invest in innovation related work and enhancing the innovation capabilities of enterprises. The influence of large models on the form of software products promotes the intelligence of software capabilities. The large model comprehensively improves the intelligence level of software from multiple dimensions such as multimodal input and output, intelligent recognition, data processing, and decision implementation. The enhancement of multimodal capabilities enables software to support information processing in multiple modalities such as text, images, and speech, and establish data associations between different modalities, providing users with a more comprehensive and flexible interactive experience. The improvement of understanding and generation capabilities enables software to no longer be limited to traditional rule-based and expert style processing methods, but to understand input information more widely and achieve knowledge understanding and application in cross domain scenarios, significantly enhancing the ability to process and generate complex information. The enhancement of decision-making ability enables software to autonomously learn, plan, and call relevant tools, and make real-time adjustments based on execution results, user behavior, and feedback, strengthening the ability to implement decisions. Drive changes in software technology. Traditional software technology will be redefined under the push of big models. At the software implementation level, large models will be deeply integrated with traditional software in various ways, such as through embedded systems, knowledge bases and RAG, or using single or multiple agents, to enhance the software's ability to understand, generate, and make decisions in various dimensions. At the software design level, based on the capability of big models, more software will develop towards AI native direction. Traditional software is driven by processes or data, while in the future, AI native software will be event driven, bringing a new user experience to software applications, making software functions more convenient, efficient, and flexible. At the same time, large models will become the technological foundation of intelligent software. In the future, providers of large model services (MaaS services) will provide more, faster, more flexible, more stable, and higher quality model services to more enterprises, driving the scale effect of software intelligence. Promote deeper integration between software and industry scenarios. The large model enhances the customization capability of software, especially the training and learning of industry specific large models on industry-specific datasets, enabling the empowered software to more effectively understand scenario requirements, apply industry knowledge more efficiently, handle scenario problems more flexibly, and generate more application scenarios. Taking the manufacturing industry as an example, compared to traditional automation software that can only execute preset programs, software with large models can analyze business needs in real time and dynamically adjust production processes, significantly improving production efficiency and product quality, and promoting the intelligent upgrading of the manufacturing industry. Inject new impetus into the development of industrial software. Under the background of technological blockade, domestic industrial software faces many challenges, especially in the field of research and development design software (such as CAD, EDA, CAE, etc.), where the domestication rate is low and the competitiveness is weak. The development of large-scale modeling technology provides new possibilities for the independent innovation and competitiveness enhancement of domestic industrial software. Firstly, it accelerates technological breakthroughs in key areas. Secondly, it helps to improve the technical capabilities and intelligent upgrading of industrial software. The impact of big models on the software industry drives the transformation of software practitioners. One is skill updating. With the development of AI technology, software industry practitioners need to constantly learn new skills and tools to adapt to the constantly changing technological environment. For example, programmers may need to learn how to collaborate with large models and even fine tune their own AI models. The second is the transformation of professional roles. The widespread application of AI may lead to a decrease in demand for certain positions, while also creating new ones. Practitioners need to adapt to these changes. For example, programmers may transform some into data annotators or prompt word engineers. The third is workflow transformation. AI technology can automate many traditional software development tasks, which requires practitioners to adapt to new workflows and pay more attention to innovation and strategic task requirements. Promote the transformation of software enterprises. In the future, software companies will face transformation pressure. On the one hand, more companies will transform to provide services related to the implementation of large models. These companies need to continuously improve their capabilities to cope with the role of large model service providers, thereby expanding the scope of software services. On the other hand, enterprises will face a transformation in their business models. As software development and customization become more accessible and inclusive, software companies may not have sustainable competitiveness by providing development and testing services to generate revenue. Instead, they may adopt a subscription or pay as you go business model. At the same time, there will be more small software companies with a few or a dozen people in the industry, which may have strong competitiveness based on emerging technologies such as big models. Traditional software companies will face challenges in how to transform to cope with more intense competition. Promote structural changes in the software industry. One is that large models will become the core content of the software industry chain, injecting new vitality and innovation into the industry, and building an important foundation for the intelligent transformation of software. The second possibility is that the demand for software outsourcing services may continue to decrease, and the demand for data annotation, prompt engineering, etc. will gradually increase, which will lead to a decrease in software outsourcing service providers and an increase in large model service and application providers. Thirdly, the toolchain used for software production will face reshaping, such as DevOps toolchain, which may become history under the impact of the powerful capabilities of generative AI. The challenge faced by the software industry in the era of big models is the lack of high-quality datasets. Code and other software related data are subject to privacy and security regulations, open and closed source protocols, and other constraints, resulting in high and complex data acquisition costs. As of September 2024, according to the TIOBE index, the current number of programming languages has exceeded 200. This leads to a lack of large datasets for model training in the industry, especially in scenarios such as embedded code in the industrial sector. In addition, the production process of high-quality datasets is not standardized enough, such as collecting, organizing, and processing code, test cases, and other data into high-quality data required for large models, which is complex and costly. Security challenges of data, models, and tools. Generative content (such as code) may bring more uncontrollable risks, so enterprises should build risk defenses from multiple dimensions of data, models, and tools to address security challenges. The first line of defense for data is the code and other datasets required for model training and optimization, which may face issues such as sensitive code data leakage, unauthorized code training, and illegal code inference; The second line of defense for large models in the inference and management stages may face issues such as brute force attacks, illegal exploitation, illegal questioning, sensitive content inference, malicious use to generate malware code or attack scripts, etc; The third line of defense is to inspect and process the input and output of intelligent development tools, and pay attention to the data transmission and integration security of associated code libraries and third-party software development tools, building the last layer of security fence to reduce application risks. The challenge of cultural cognition and talent structure adjustment. The intelligent transformation faces challenges in cultural reshaping, cognitive enhancement, and talent structure optimization. One is the reshaping of organizational culture, which requires organizations to have a cultural atmosphere of open cooperation, continuous learning, and innovation, promote cross departmental communication and collaboration, and break down information silos; The second is the improvement of the cognitive level of all employees. From top to bottom, every employee should have a deep understanding of the potential and challenges of intelligence, objectively and correctly understand artificial intelligence technologies such as big models and their roles; The third is the adjustment and supplementation of talent structure. As the process of data governance and model training optimization requires the participation of professional technical personnel, supplementing talents or adjusting or integrating the structure of AI teams and software engineering teams can match the needs of capacity building and application. In the future, software and its production processes will become more intelligent, including intelligent and diversified interactive forms, autonomous and low threshold research and development paradigms. AI applications will be ubiquitous, and software development will also be ubiquitous. As we are in the wave of the times, we must embrace AI and big models, explore the possibility of new technologies landing, and lead the software industry to accelerate on the track of intelligence with innovative thinking; Secondly, we must be down-to-earth and never forget our original intention, deeply explore the deep-seated problems of software engineering, reasonably layout the goals of improving quality, reducing costs, and increasing efficiency, and promote the sustainable and prosperous development of the software industry. (New Society)