The fourth session of the Zhejiang University DeepSeek Series Special Online Open Course, lectured by Zhu Chaoyang and Chen Jianhai. "The Fire of DeepSeek Can Start a Prairie Fire" exploresDeepSeekthe combination of technology and Chinese cultural genes, analyzes the advantages in artificial intelligence development, such as the efficiency of Chinese characters, the innovativeness of knowledge distillation technology, etc., and discusses AI ethics and security issues. "DeepSeek's Local Deployment and the Future of AI General Education" focuses on DeepSeek's local deployment and the future of AI general education, introduces DeepSeek's performance, application scenarios, and practices in the education field, and emphasizes its empowering role in artificial intelligence education at universities. It showcases DeepSeek's exploration and practice in technological breakthroughs and educational applications, providing new ideas and directions for the localized development of artificial intelligence and the popularization of education.
Obtain"The Fire of DeepSeek Can Start a Prairie Fire"
"DeepSeek's Local Deployment and the Future of AI General Education"
Original PDF file, scan the QR code to follow and reply: 20250221
"The Fire of DeepSeek Can Start a Prairie Fire"
"The Fire of DeepSeek Can Start a Prairie Fire" is the speech content of Professor Zhu Chaoyang of Zhejiang University in the fourth session of the "DeepSeek Series Special Online Open Course." It mainly explores the development, application, and connection of artificial intelligence (especiallyDeepSeektechnology) with Chinese cultural genes. Starting from the spirit of openness, inclusiveness, and sharing in Chinese culture, Professor Zhu analyzes DeepSeek's advantages in technological breakthroughs, corpus processing, model training, and other aspects, especially its innovation in Chinese corpus processing. The article discusses issues such as AI ethics and security and algorithmic bias, and proposes solutions to enhance local DeepSeek capabilities, such as fine-tuning and retrieval-augmented generation (RAG). Finally, it looks ahead to the development path of Chinese AI, emphasizing the importance of combining Chinese cultural genes with technological innovation.
Understanding the Path to DeepSeek's Success from the Perspective of Chinese Cultural Genes
- Seek benevolence and obtain benevolence, seek Buddha and obtain Buddha: Based on the story of Tang Monk in "Journey to the West," this leads to reflection on the goals and values of artificial intelligence.
- Openness, inclusiveness, and sharing have always been virtues of our Chinese nation: Emphasizes the positive significance of the virtues of openness, inclusiveness, and sharing in Chinese culture for the development of artificial intelligence.
- The English Tower of Babel curse: Based on the story of the Tower of Babel in the Bible, this draws an analogy to the communication and division problems English may face in AI development.
- English will become a huge challenge for AI development: Analyzes the challenges of English in AI development, including issues in vocabulary size, storage efficiency, and other aspects.
- The inefficiency and wasted space of English storage: Uses specific examples to demonstrate the inefficiency and wasted space of English storage.
- The efficiency and inheritable nature of Chinese: Discusses the advantages of Chinese in AI training, including its efficiency and cultural continuity.
- Comparison of training corpora between DeepSeek and ChatGPT: Compares the differences between DeepSeek and ChatGPT in training corpora, emphasizing DeepSeek's advantages in Chinese corpus processing.
- Using Chinese characters for model training will win in this AI competition: Explores the potential advantages of using Chinese characters for model training in AI competition.
- The paradigm of achieving miracles through massive effort: Discusses methods of improving AI performance by increasing parameters, data, and computing power.
- The indomitable spirit of the Chinese nation: Emphasize the role of the Chinese nation's unyielding spirit in promoting DeepSeek's technological breakthrough.
- DeepSeek's Technological Breakthrough: Introduce DeepSeek's technological innovations and breakthroughs, including computing power optimization, algorithm innovation, and architecture innovation.
Two Controversies Surrounding DeepSeek
- Controversy One: Is Knowledge Distillation Plagiarism or Innovation: Discuss in detail the controversy over whether knowledge distillation technology constitutes plagiarism, as well as its innovativeness.
- Controversy Two: Is Artificial Intelligence a Resource Black Hole or an Efficiency Revolution: Explore the impact of AI on resource consumption and efficiency improvement.
Reflections on the Future Development Path of Artificial Intelligence
- The Dao and Shu of AI: Drawing on the ideas of the Dao De Jing, discuss the relationship between AI's "Dao" (laws, origins) and "Shu" (methods, techniques).
- AI's Ethical Security and Moral Dilemmas: Discuss issues related to AI in ethical security and moral dilemmas, such as the trolley problem in autonomous driving.
- AI's Algorithmic Bias and Dominance: Explore AI algorithmic bias and its impact on society and decision-making.
DeepSeek Private Deployment Application Methods
- Two Solutions for Enhancing Local DeepSeek Capabilities: Introduce two methods for enhancing local DeepSeek capabilities: fine-tuning and retrieval-augmented generation (RAG).
- Local Deployment Retrieval Augmentation-RAG: Introduce RAG technology and its application in local deployment in detail.
- RAG Deployment Process: Introduce the RAG deployment process step by step, including download and installation, interface configuration, model and knowledge base management, etc.
- Security Report on ollama: Provide a report on ollama security risks and hardening recommendations.
- One of Two Techniques for Large Model Fine-Tuning: LoRA: Introduce LoRA technology and its application in large model fine-tuning.
- One of Two Techniques for Large Model Fine-Tuning-Backbone: Introduce Backbone fine-tuning technology and its application scenarios.
- Low-Code Fine-Tuning Method-Based on LLaMA-Factory: Introduce in detail the low-code fine-tuning method based on LLaMA-Factory, including installation and deployment, training template creation, and the training process.
- A Simple Application of DeepSeek Local Fine-Tuning: Showcase applications of DeepSeek local fine-tuning in fields such as AI intelligent teaching assistants.
- Integration of the Intelligent Teaching Assistant System and the Virtual Simulation Laboratory: Discuss the integrated application of the intelligent teaching assistant system and the virtual simulation laboratory.
- Looking Ahead to a Path of AI That Belongs to China: An outlook on the future development path of AI in China, emphasizing the importance of combining Chinese cultural genes with technological innovation.
"The Localized Deployment of DeepSeek and the Future of AI General Education"
"The Localized Deployment of DeepSeek and the Future of AI General Education" is the content of a lecture by Professor Chen Jianhai of Zhejiang University in the fourth session of the "DeepSeek Series Special Online Open Course." It mainly explores the localized deployment strategies of the DeepSeek model and the potential impact and application prospects in promoting artificial intelligence (AI) general education. It introduces how DeepSeek, as an advanced AI model, can meet the needs of different regions and industries based on localized deployment and improve the efficiency and security of data processing. It explains the importance of AI general education and the application of DeepSeek in the field of education. It discusses how DeepSeek can help educators and students better understand and apply AI technology, bringing innovation to future education models.
The New Era Background of DeepSeek AI
- Internet big data triggers intelligentization: Introduces the development process from the agricultural era to the internet era, the big data era, and then to the new era of artificial intelligence, emphasizing the intelligent transformation brought about by the information technology revolution.
- The rise of generative artificial intelligence (AIGC): Showcases a video generated by the LumaAI large model DreamMachine, explaining the applications of generative artificial intelligence in text-to-text, text-to-image, text-to-video, and other areas.
- Past, present, and future—the era of human-machine coexistence: Describes the transition from the real world of the past to today's digital avatars, and then to the future metaverse world, emphasizing the arrival of the era of human-machine coexistence.
- The era where artificial intelligence (AI) is everywhere: Lists applications of artificial intelligence in various fields, such as security, customer service, intelligent assistants, business process automation, intelligent image understanding, precision marketing, machine translation, intelligent search ranking, intelligent travel, autonomous driving, intelligent logistics, financial applications, medical applications, education and learning, and more.
- The origin and definition of artificial intelligence: Introduces the origin of artificial intelligence (the 1956 Dartmouth Conference), mentions that intelligence is the general term for wisdom and ability, and Howard Gardner's theory of multiple intelligences, including linguistic intelligence, musical intelligence, logical intelligence, spatial intelligence, bodily-kinesthetic intelligence, human-machine intelligence, cognitive intelligence, and more.
- From human intelligence to artificial intelligence: Cites the explanation of “wisdom” and “ability” in the “Correcting Names” chapter of Xunzi, expounds the relationship between human intelligence and artificial intelligence, and emphasizes that artificial intelligence uses machines as its carrier and uses artificial methods and technologies to simulate and realize human intelligence.
- DeepSeek is here, who is it: Introduces Hangzhou DeepSeek AI Basic Technology Research Co., Ltd. and its founder Liang Wenfeng, and DeepSeek as a high-performance, low-cost domestic open-source large model.
- What is DeepSeek: Describes the DeepSeek large model product series, including DeepSeek-Math, DeepSeek-Coder, DeepSeek-VL, V2, V2.5, V3, R1, R1-Zero, and more.
- DeepSeek's earth-shaking achievements: Summarizes DeepSeek's important development milestones, such as model releases, deployment on the national supercomputing platform, open-sourcing, and more.
- Who is DeepSeek? Let's put it in plain terms: Explains DeepSeek's different versions based on metaphors, such as the distilled version and the full version, and their training foundations.
- DeepSeek breaks out of the circle and sweeps the globe: Describes DeepSeek's rapid growth and user registration situation worldwide.
- DeepSeek is fully integrated across multiple scenarios: Describes DeepSeek's applications in multiple industries and fields, and its cooperation with world-class cloud computing giants.
- What DeepSeek can do (what ChatGPT can do, DeepSeek can do too): Lists the application scenarios supported by DeepSeek, such as intelligent dialogue, text generation, semantic understanding, and more, emphasizing its versatility.
DeepSeek's localized deployment practice
- What is DeepSeek's localized deployment: Explains the definition of DeepSeek's localized deployment, deploying artificial intelligence systems on user-specified local servers or infrastructure to achieve data sovereignty control, cultural adaptation optimization, and compliance assurance.
- Advantages of localized deployment: Introduces in detail the advantages of localized deployment, including data privacy and security, legal compliance, performance optimization and low latency, a high degree of customization, stability and reliability, long-term cost-effectiveness, technological autonomy, and more.
- Simple use of DeepSeek's official website connection: Provides the connection address for the DeepSeek official website and introduces the steps for registering an account for first-time use.
- Local run based on Ollama (macOS or Linux): Introduces methods for installing Ollama and running DeepSeek on macOS or Linux systems, including installing the Ollama package and running different versions of the DeepSeek model.
- Preliminary preparation for DeepSeek deployment - machine environment: Explains the requirements for personal use and organizational use in terms of machine configuration, operating system, and software.
- Localized deployment of Ollama - download and installation: Provides the download address of the Ollama official website and shows the download and installation steps.
- Installation of DeepSeek local service: Describes the process of running the DeepSeek local service after installation is complete.
- Local running and use of the DeepSeek model - ollama [command-line run on Windows]: Introduces the method for running the DeepSeek model based on the command line on Windows systems.
- DeepSeek local version - based on PageAssist for Ollama (web version): Introduces the functions and usage of PageAssist for Ollama, including its official website entry URL and the address for downloading the plugin.
The Future of DeepSeek and AI General Education
- National strategy calls for the "Artificial Intelligence +" action: Mentions that the 2024 Government Work Report explicitly mentioned the "Artificial Intelligence +" action for the first time, emphasizes the important position of AI technology in the national strategy, and describes the situation in which universities are successively advancing artificial intelligence general education.
- Zhejiang University establishes the Artificial Intelligence Education and Teaching Research Center: Introduces the background and goals of Zhejiang University's establishment of the Artificial Intelligence Education and Teaching Research Center, including building a public course system for undergraduate and graduate AI courses, publishing a series of high-level AI textbooks, creating a series of AI talent training projects, and forming an interdisciplinary high-level teaching team.
- Zhejiang University forms an interdisciplinary high-level artificial intelligence teaching team: Describes the situation of Zhejiang University forming an interdisciplinary teaching team, and realizing a university-wide linkage and multi-department collaborative work mechanism, forming a new integrated ecosystem of "Artificial Intelligence +".
- Zhejiang University builds an artificial intelligence talent cultivation course system: Introduces Zhejiang University's construction of core courses for the artificial intelligence major, opening artificial intelligence micro-majors, offering interdisciplinary artificial intelligence courses, and offering multi-level basic artificial intelligence general education courses.
- Reform the training program for general education courses in the computer category: Shows the specific course settings of Zhejiang University's reform of the training program for general education courses in the computer category, including course number, course name, credits, weekly class hours, recommended semester, and course description.
- Zhejiang University artificial intelligence basic general education course textbook system: Explains the artificial intelligence basic general education course textbook system constructed by Zhejiang University, including course settings, credits, course introductions, etc. for students in different majors.
- Course offering status of instructors: Introduces the course offerings of Zhejiang University's artificial intelligence general education courses in the 2024-2025 fall and winter semester and the 2025 early summer semester, including course names, target majors, credits, semesters, course introductions, number of students selecting the course, parallel classes, main instructors, etc.
- Adopts the self-compiled textbook Artificial Intelligence General Education Basics (Science, Engineering, Agriculture, and Medicine): Mentions the textbook's authors, publisher, publication time, and the textbook's accompanying syllabus, curriculum system, cases, PPT materials, and other digital resources.
- Foundations of Artificial Intelligence A – A General Education Foundation Course for Science, Engineering, Agriculture, and Medicine: Provides a detailed introduction to the course code, Chinese name, English name, credit hours, prerequisites, target audience, teaching objectives, etc.
- Four Unified Course and Grade Assessment: Explains the "four unified" requirements for university-wide general education courses, including unified syllabus, unified requirements, unified textbook, unified examination, and the specific scheme for grade distribution.
- Syllabus – Learning Objectives: Describes the course's learning objectives, including knowledge memorization, abilities, hands-on skills, innovation, technology, ideological and political education, etc.
- Syllabus – Educational Objectives: Expounds the course's educational objectives, including cultivating innovative thinking, critical thinking ability, interdisciplinary integration awareness, collaboration and management skills, ethics and social responsibility, international perspective, etc.
- Syllabus – Course Knowledge Point System: Lists in detail the course's knowledge point system, including the specific knowledge points and class hour arrangements for the Basic Introduction section, Machine Learning section, Deep Learning section, and Large Language Model section.
- Syllabus – Teaching Schedule: Displays the course's teaching schedule, including detailed information such as number, module, knowledge points, and class hours.
- Summary of How to Teach and How to Learn: Introduces learning methods and teaching logic, emphasizing the learning path of basic introduction, machine learning, deep learning, and large language models, as well as learning methods such as combining classroom listening with extracurricular independent learning and valuing practice.
- DeepSeek's Localized Deployment Helps Zhejiang University's "Zhejiang University Mr." New Ecosystem Empower General Education: Describes how Zhejiang University, centering on "student growth," builds the "Mr." intelligent agent application portal, constructs a new user ecosystem for intelligent application services in universities, and the role of DeepSeek's localized deployment in it.
- Case: Zhejiang University – Mr. Connects to DeepSeek: Introduces the functions and features of Zhejiang University's "Mr." intelligent agent application portal, including intelligent Q&A, intelligent agent development, teaching and training, etc.
- Zhejiang University Mr.'s Technical Architecture: Displays the positioning, approach, scenarios, platform, facilities, and other aspects of the "Mr." technical architecture.
- DeepSeek Empowers AI General Education: Explains the specific content of DeepSeek's empowerment of AI general education in the Basic Introduction section, Machine Learning section, Deep Learning section, and Large Language Model section, including textbook content, textbook cases, reducing teachers' burden and professional growth, cultivating students' literacy and creativity, reconstructing the educational ecosystem, etc.
- Zhejiang University General Education Courses Support the Construction of "Famous Teachers and Famous Courses" for Artificial Intelligence General Education in Undergraduate Universities in Zhejiang Province: Introduces a batch of research-level high-quality online course resources led by Zhejiang University, including course names, levels, target participants, credit recommendations, leading universities/experts, and other information.
Obtain"The Fire of DeepSeek Can Start a Prairie Fire"
"DeepSeek's Localized Deployment and the Future of AI General Education"
PDF Original File, scan the QR code to follow and reply: 20250221