脈絡拆解: AWS re:Invent 2022 Swami Sivasubramanian Keynote

Post Title Image

tl;dr 重點摘要

  • SageMaker 演進:整合分析、機器學習和生成式 AI 能力的全新統一體驗
  • Bedrock 強化:新的模型合作夥伴(Luma AI、Poolside)、提示最佳化功能,以及進階 RAG 功能
  • Amazon Q 改進:機器學習模型開發輔助、商業情境分析,以及開發者生產力工具
  • 基礎架構最佳化:Hyperpod 任務治理、彈性訓練計畫,以及運算資源管理
  • 安全性和負責任的 AI:強化多模態毒性偵測和自動推理檢查


知識圖譜

graph LR subgraph Core["AWS Core Services"] SM["SageMaker"] BR["Bedrock"] AQ["Amazon Q"] NP["Neptune"] end subgraph SageMaker["SageMaker Features"] subgraph Training["Training & Inference"] HPod["Hyperpod"] FlexPlan["Flexible Plans"] TaskGov["Task Governance"] end subgraph Studio["Studio Features"] Canvas["Canvas with Q"] MLOps["MLOps"] Collab["Collaboration"] end end subgraph Bedrock["Bedrock Features"] subgraph Models["Foundation Models"] subgraph AWSModels["AWS Models"] Nova["Nova Family"] NovaMulti["Multimodal"] NovaVid["Video/Image"] NovaText["Text-to-Text"] end subgraph Partners["Partner Models"] Anthro["Anthropic/Claude"] Meta["Meta/Llama"] Mistral["Mistral AI"] Stability["Stability AI"] Luma["Luma AI"] Pool["Poolside"] end end subgraph Safety["Safety Features"] Guard["Guardrails"] Toxic["Toxicity Detection"] Reason["Automated Reasoning"] end subgraph Knowledge["Knowledge Features"] RAG["RAG"] GraphRAG["Graph RAG"] StructRAG["Structured Data RAG"] DataAuto["Data Automation"] Kendra["Kendra GenAI"] end end subgraph AmazonQ["Amazon Q Features"] subgraph Dev["Developer Tools"] Code["Code Gen"] Debug["Debugging"] Doc["Documentation"] end subgraph Analytics["Analytics"] QS["QuickSight"] Scen["Scenarios"] NLQ["Natural Queries"] Dash["Dashboards"] end subgraph Business["Business Tools"] Search["Enterprise Search"] Sum["Summaries"] Content["Content Generation"] end end %% Core Connections SM --> Training SM --> Studio BR --> Models BR --> Knowledge BR --> Safety AQ --> Dev AQ --> Analytics AQ --> Business %% Neptune Integration NP --> GraphRAG %% Model Relationships Nova --> NovaMulti Nova --> NovaVid Nova --> NovaText %% Feature Connections HPod --> FlexPlan HPod --> TaskGov
Loading comments…