4.3 KiB
title, type, source-type, category, tags, date-added, video-source, audio-source, status
| title | type | source-type | category | tags | date-added | video-source | audio-source | status | |||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Public Cloud Learning Sessions-Introduction to Artificial Intelligence (AI) Machine Learning (ML) - 20240206 160153-Meeting Recording | cloud-learning | video | DevOps & SRE/09_Serverless_AI |
|
2026-04-14 | nas:///volume2/work/Public Cloud Learning Sessions/Public Cloud Learning Sessions-Introduction to Artificial Intelligence (AI) Machine Learning (ML) - 20240206_160153-Meeting Recording.mp4 | summarized (Gemini 摘要) |
Public Cloud Learning Sessions-Introduction to Artificial Intelligence (AI) Machine Learning (ML) - 20240206 160153-Meeting Recording
Source: NAS /volume2/work/Public Cloud Learning Sessions/Public Cloud Learning Sessions-Introduction to Artificial Intelligence (AI) Machine Learning (ML) - 20240206_160153-Meeting Recording.mp4
Type: VIDEO | Category: 09_Serverless_AI
Status: 🟡 Awaiting Whisper transcription → Summary
Introduction to AI/ML with AWS
The session introduces AI/ML, generative AI in AWS, and implementing the data science lifecycle in AWS, presented by Suraav Paul, AWS Senior Solutions Architect. It covers how AI/ML is transforming businesses and how Amazon and AWS are accelerating this transformation.
AI replicates tasks needing human intelligence, often seeking probabilistic outcomes via machine learning, which uses data to create decision logic or models. Classification AI identifies patterns, predictive AI forecasts trends, and generative AI creates content using foundation models (FMs). Amazon has invested in ML for 20 years, using it for recommendations, robotics, forecasting, and Alexa.
AWS helps customers use AI in four areas: enhancing customer experiences, enabling better decisions, improving operations, and creating new products. The AI Use Case Explorer helps customers find relevant AI use cases. Responsible AI includes fairness, explainability, robustness, governance, transparency, and privacy/security. AWS offers pre-built algorithms, models, and solutions, democratizing access to AI/ML with tools like Amazon SageMaker Canvas.
We believe most customer experiences and applications will be reinvented with generative AI, powered by foundation models with billions of parameters.
Amazon Bedrock is a fully managed service for building and scaling generative AI applications with foundation models, allowing customization with proprietary data while maintaining security and privacy. Bedrock offers access to various FMs, including Amazon Titan models, which provide capabilities, competitive pricing, and strong performance. Data customization techniques include fine-tuning (using labeled datasets) and continued pre-training (using unlabeled datasets). Retrieval augmented generation (RAG) fetches data from company sources for relevant responses. Agents for Amazon Bedrock plan and execute multi-step tasks using company systems and data sources. Guardrails for Amazon Bedrock enables safeguards tailored to application requirements and responsible AI policies.
ML Ops combines machine learning and operations, involving people, technology, and processes for collaborative ML solutions. It requires a diverse team and a culture that encourages collaboration. The ML Ops process includes data, training, and inference pipelines. The data pipeline involves data collection, integration, and preparation using services like Amazon S3 and Amazon Redshift. The training pipeline focuses on feature engineering, model training, and hyperparameter tuning using SageMaker. The inference pipeline deploys and monitors models using SageMaker's real-time endpoint. ML Ops addresses concerns around data provenance, model management, and deployment workflows, in addition to DevOps practices like CI/CD and monitoring.
During the Q&A, it was clarified that training data used from a company won't be used generally by the model. For first-party models, no additional licensing is needed, but third-party models have their own licensing agreements. Bedrock stores data only for the request-response cycle, ensuring data privacy. Previously trained models can be imported to SageMaker, while Bedrock offers a managed environment without needing to manage deployment infrastructure.
AI is a way to describe any system that can replicate tasks that previously required human intelligence.