当前位置:首页 > 报告详情

从提示到微调:利用人工智能和LLM技术服务于公用事业领域 [LRN1133].pdf

上传人: Fl****zo 编号:971426 2025-11-08 50页 7.59MB

1、 From Prompts to Fine Tuning(LRN1133)Harnessing AI&LLMs for the Utility SectorAshish AgarwalSr.Director,Oracle Utilities Research LabOctober 2025About Me30 years of industry experienceTelecommunication&Utility industryLeading the Industry Innovation labApplying AI to Industry domainsMulti-Modal AI S

2、oftware DeveloperHow about you?Copyright 2025,Oracle and/or its affiliates|Confidential:Internal/Restricted/Highly Restricted3Agenda4Copyright 2025,Oracle and/or its affiliates|Confidential:Internal/Restricted/Highly Restricted1.Prompt Engineering2.Retrieval Augmented Generation(RAG)3.Agents4.Fine T

3、uning5.Demos 6.Quiz5Copyright 2025,Oracle and/or its affiliates|Confidential:Internal/Restricted/Highly RestrictedProcedural.Programs for algorithms.What.Not how.Structured dataMultimodal unstructured dataSpecialized programming skillsDemocratizedPredictableCan be“creative”(temperature)Automation of

4、 routine menial tasksBetter then average human at complex cognitive tasksTask Oriented Mastery of all human knowledgeTraditional ComputingGenerative AI Generative AI What makes it different?Copyright 2025,Oracle and/or its affiliates|Confidential:Internal/Restricted/Highly Restricted6RAGAnswer Quest

5、ionse.g:Chatbot answering FAQsAgentsAction/Reasoninge.g:Recommend Best ActionTuningAdapt LLM to a specific domaine.g:Chatbot with a personalityLLM Interaction PatternsAuthor Contente.g:Summarize this call transcriptPrompt EngineeringLLMTextTextLLMTextTextDataLLMTextTextDataDB ToolApp ToolWeb ToolLLM

6、TextTextPrompt EngineeringPrompt Engineering8Copyright 2025,Oracle and/or its affiliates|Confidential:Internal/Restricted/Highly Restricted Large Language Models=Smart internsLet us explain.The best team possible provides the best solutions for real problems9Copyright 2025,Oracle and/or its affiliat

word格式文档无特别注明外均可编辑修改,预览文件经过压缩,下载原文更清晰!
三个皮匠报告文库所有资源均是客户上传分享,仅供网友学习交流,未经上传用户书面授权,请勿作商用。
根据报告的内容,本文主要探讨了人工智能(AI)和大型语言模型(LLM)在公用事业领域的应用。以下是全文关键点: 1. **AI与LLM特点**:AI自动化常规任务,优于人类在复杂认知任务上;LLM作为“智能实习生”,提供最佳解决方案。 2. **Prompt Engineering**:利用LLM进行内容生成、分析、合规和测试数据生成。 3. **Retrieval Augmented Generation (RAG)**:通过结合LLM和私有企业数据,减少LLM的幻觉,提供数据来源引用。 4. **Agent**:LLM代理可执行复杂任务,具备推理、记忆和协作能力。 5. **Fine Tuning**:调整LLM以适应特定领域,包括定制输出、新个性和领域采用。 6. **案例演示**:展示了如何使用Oracle AI工具进行呼叫转录、情感分析和知识总结。
"RAG如何提升AI效率?" "AI助手如何解决复杂问题?" "LLM微调在行业应用中如何体现?"
客服
商务合作
小程序
服务号
折叠