Xiamen University: Free Download of the 120-Page PPT "DeepSeek Large Model Empowering University Teaching and Research" 192

"DeepSeek Large Models Empowering University Teaching and Research" is a popular science report produced by the Big Data Teaching Team of Xiamen University on the application of the DeepSeek large model in university teaching and research, presented by Associate Professor Lin Ziyu, and released on February 25, 2025. "DeepSeek Large Models Empowering University Teaching and Research" provides a detailed introduction to the development history of artificial intelligence, the technical principles of large models, the characteristics of the DeepSeek large model, and the practice of its local deployment in universities. It also discusses the applications of AI-generated content (AIGC), intelligent agents based on large models, and how AI empowers university scientific research and teaching.

Baidu Netdisk Download Link:https://pan.baidu.com/s/1i0P1mYxBYpgl23OGC5EDxQ?pwd=6aer Extraction Code: 6aer

I. A Brief History of Artificial Intelligence

• Turing Test: In 1950, Alan Turing proposed a test method for machine intelligence, laying the foundational concept of artificial intelligence.

• Birth of Artificial Intelligence: In 1956, the Dartmouth Conference marked the formal birth of artificial intelligence as a discipline.

• Stages of Development: Artificial intelligence has gone through multiple stages from its birth to the present, including the budding period, the sedimentation period, and the explosion period.

• Future Outlook: Artificial intelligence may go through five stages in the future, gradually developing toward Artificial General Intelligence (AGI).

II. AI Thinking

• Understanding and Distinguishing: Everyone should understand the basic operating modes of artificial intelligence and develop the mindset to distinguish between human capabilities and machine capabilities.

• Collaboration Skills: Learn to collaborate with artificial intelligence and understand how to use AI tools.

III. Large Models: The Frontier of AI

• Concept of Large Models: Large models are models based on deep learning technology, featuring massive parameters, powerful learning capabilities, and generalization abilities.

• Development History: Large models have gone through three stages—budding, sedimentation, and explosion—and the demand for computing power has also been evolving.

• Classification: Large models can be divided into language large models, visual large models, and multimodal large models, and can also be classified by application domain into general large models (L0), industry large models (L1), and vertical large models (L2).

• Reasoning Large Models: For example, OpenAI's o3, which can perform complex logical reasoning and is suitable for solving multi-step problems.

• Technical Principles: Based on the Transformer architecture, it achieves powerful language understanding and generation capabilities through encoder-decoders and self-attention mechanisms.

IV. Local Deployment of DeepSeek Large Models in Universities

• Introduction to DeepSeek: DeepSeek is a large model with outstanding performance, particularly excelling in mathematical reasoning, with low training costs.

• University Practice: Several universities, such as Zhejiang University and Beijing Normal University, have begun exploring the local deployment of DeepSeek for teaching and research.

• Deployment Advantages: Local deployment enables offline efficient use, customization, data privacy protection, cost optimization, and more.

• Deployment Solutions: Including the combination of model fine-tuning and local knowledge base (RAG) to enhance the model's performance in specific domains.

• Technical Challenges: Universities face issues such as insufficient computing power and complex technology stacks, but Tsinghua University's KTransformers project has achieved a breakthrough in low-cost deployment.

V. AIGC Application and Practice

• Overview of AIGC: AIGC stands for Artificial Intelligence Generated Content, which uses AI technology to generate text, images, audio, video, and more.

• Application Scenarios: Covering e-commerce, office work, gaming, entertainment, film and television, art, education, design, media, and many other fields.

• Tools and Prompts: Introduces various AIGC tools (such as ChatGPT, DeepSeek, Tongyi Qianwen, etc.) and tips for using prompts.

• Practical Cases: Including text generation, image generation, voice generation, video generation, assisted programming, AI search, and intelligent office work.

VI. Intelligent Agents Based on Large Models

• Concept of Intelligent Agents: Intelligent agents are systems that mimic human intelligent behavior, capable of perceiving the environment and making autonomous decisions.

• Intelligent Agents Based on Large Models: Built using large language models, they possess autonomy, interactivity, and adaptability.

• Application Scenarios: Intelligent agents can be applied in multiple university scenarios, including teaching, management, scientific research, and services.

• Case Study: For example, the super-intelligent agent of Shanghai Maritime University achieves learning assistance, teaching assistance, research assistance, management assistance, and international exchange assistance.

VII. AI Empowering University Scientific Research

• Improvement in Research Efficiency: DeepSeek helps researchers quickly complete literature retrieval, data analysis, experimental design, and paper writing.

• Interdisciplinary Support: Breaking down disciplinary barriers and providing assistance across multiple fields.

• Success Case: An Italian research team used DeepSeek to publish an SCI paper within one day.

• Notes for Attention: Be vigilant about the accuracy of AI-generated content, and avoid overly strong "AI flavor" that leads to paper plagiarism detection failures.

VIII. AI Empowering University Teaching

• Innovative Teaching Models: Including adaptive learning systems, AI teaching assistants, teacher AI assistants, intelligent educational affairs systems, and interdisciplinary cultivation.

• Improvement in Teaching Effectiveness: Significantly enhancing student learning outcomes through intelligent learning companions, personalized learning paths, and multimodal teaching resources.

• Transformation of the Teacher's Role: Shifting from a "lecturer" to a "guide," "collaborator," and "emotional connector."

• Future Outlook: AI is the "intelligent scaffolding" of education. Teachers should collaborate with AI, leveraging the combined strengths of human warmth and technological precision.

"DeepSeek Large Models Empowering University Teaching and Research" emphasizes that large models are an important research direction in the field of artificial intelligence. Their powerful language understanding and generation capabilities make them widely applicable in natural language processing, machine translation, intelligent customer service, and other fields. With the continuous development of technology, the application scenarios of large models are also constantly expanding, and they will be more widely applied in various fields of university teaching and research in the future.