AI-900 Azure AI Fundamentals Lecture Notes
Introduction to Cloud-based AI
Speaker: Dr Amir Pourabdollah
Key Topics:
- Cloud and cloud services
- Benefits and challenges of cloud-based AI
- Principles of responsible AI
Sources:
- Microsoft Azure AI Fundamentals (AI-900)
- Microsoft Azure Fundamentals: Describe cloud concepts
Cloud Computing Overview
What is Cloud Computing?
- Delivery of computing services over the internet
- Enables faster innovation, flexible resources, and economies of scale
Benefits of Cloud Computing
- High availability and scalability
- Global reach and agility
- Disaster recovery and fault tolerance
- Elasticity and customer latency capabilities
- Predictive cost considerations and security
Challenges of Cloud Computing
- Connectivity and data privacy concerns
- Black-boxed solutions and costs
- Data security concerns
- Quality of Service (QoS) and technology familiarity
Types of Cloud Environments
Public Cloud
- Owned by cloud service providers
- Provides resources to multiple organizations and users
- Accessed over a secure network (internet)
Private Cloud
- Created by organizations in their datacenters
- Organization operates the services
- No external user access
Hybrid Cloud
- Combines public and private clouds
- Applications run in the most appropriate location
Cloud Service Models
Infrastructure as a Service (IaaS)
- Pay-as-you-go IT infrastructure
- Rent servers, VMs, storage, networks, and OS from providers
Platform as a Service (PaaS)
- Environment for building, testing, and deploying applications
- No need to manage underlying infrastructure
Software as a Service (SaaS)
- Use cloud-based apps over the internet (e.g., Microsoft Office 365)
Why Cloud-based AI?
- Combines ML/AI with cloud computing
- Knowledge and resource sharing
- Offloading computation from client-side devices
- Facilitates growth in IoT and embedded AI
Common Cloud Service Providers for AI
- Various comparison reviews available
Key AI Workloads
- Machine Learning and predictive models
- Anomaly Detection
- Computer Vision
- Natural Language Processing (NLP)
- Knowledge Mining
- Generative AI
Examples of Cloud-based AI Applications
Cloud-based Computer Vision
- Image classification and object detection
- Semantic segmentation and image analysis
- Optical character recognition (OCR)
Anomaly Detection
- Identify events or observations differing significantly from data patterns
- Benefits include prompt troubleshooting
Machine Learning on Cloud
- Resource-intensive traditional model development
- Cloud resources for training, optimization, and running models
- Approaches: Pre-trained models, visual interfaces, server-side programming, hybrid
Azure ML
- Repeating experiments to choose the best model
- Deployed models can be accessed via endpoint and key
Pipelining in Azure ML
- Visual tool for creating ML pipelines
- Training and inference pipelines
NLP Capabilities
- Language detection, key phrase extraction, entity detection
- Sentiment analysis, question answering, conversational understanding
- Speech to text, text to speech, and translation
Developing a Cloud-based Chatbot
- Knowledge base with Q&A pairs
- Training model for answering unseen questions
- Deploy as cloud service
Challenges and Risks with AI
Common Issues
- Bias affecting results
- Errors causing harm
- Data exposure
- Solutions not working for everyone
- Trust in complex systems
- Liability in AI-driven decisions
Principles of Responsible AI
Fairness
- AI should treat all people fairly
- No bias in ML models for applications like loan approvals
Reliability & Safety
- AI systems must perform reliably and safely
- Rigorous testing for applications like autonomous vehicles
Privacy & Security
- AI systems should ensure data privacy and security
Inclusiveness
- Empowerment and engagement of all people
Transparency
- Users should understand AI system purposes and limitations
Accountability
- Clear governance and ethical standards for AI design and deployment
Resources:
Note: This is a summary of the key points from the AI-900 Azure AI Fundamentals lecture. Use this as a reference for understanding the fundamental concepts and principles of cloud-based AI and responsible AI practices.