Fundamental AI Concepts Flashcards
Key Workloads of AI
- Machine Learning
- Computer Vision
- NLP (Natural Language Processing)
- Document Intelligence
- Knowledge Mining
- Generative AI
Azure Machine Learning Studio
- automated Machine Learning
- Azure Machine Learning Designer (graphical, no-code)
- Data Metric Visualization
- Notebooks (Jupyter)
Computer Vision Tasks
Interpret the world visually through cameras, video, and images
- Image Classification (by content)
- Object Detection (identify and classify single object in image, bounding box)
- Semantic Segmentation (classify individual pixels)
- Image Analysis (extract info from images)
- Face Detection, Analysis, Recognition
- OCR (Optical Character Recognition) (detect and read text in images)
- Azure (AI) Vision Studio
Computer Vision: Image classification
Training a machine learning model to classify images based on their contents
Computer Vision: Object Detection
Classify individual objects within an image & identify their location with a bounding box
Computer Vision: Semantic Segmentation
Individual pixels in the image are classified according to the object to which they belong
Computer Vision: Image Analysis
Extract information from images, including “tags” that could help catalog the image
Computer Vision: Face detection, analysis and recognition
A specialized form of object detection that locates human faces in an image
Computer Vision:
Azure (AI) Vision Studio
- Image Analysis
- Face Detection and recognition
- OCR (printed or handwritten)
NLP (Natural Language Processing)
Interpret written or spoken language and respond in kind
- analyze and interpret written text
- interpret spoken language
- synthesize speech responses
- translate to other languages
- interpret commands and determine appropriate reactions
Document Intelligence
Managing, processing, and using high volumes of data found in forms and documents
- Data collection from scanned documents
- Enhance data-driven strategies
- Enrich document search capabilities
- Azure Document Intelligence Studio
Knowledge Mining
Extract information from large volumes of often unstructured data to create a searchable knowledge store
Azure AI Search
Azure AI Search
Knowledge Mining Tool
- building internal or public indexes
- combines image processing, document intelligence and NLP to extract data
- quickly index previously unsearchable documents and extract and surface insights from large amounts of data
Generative AI
Create original content in various formats
- input natural language
- return appropriate responses in various formats, including natural language, images, code, and audio
-Azure Open AI studio
NLP in Microsoft Azure (Azure Speech Studio)
AZURE AI LANGUAGE
- understanding and analyzing text
- training conversational language models
AZURE AI SPEECH
- speech recognition and synthesis
- real-time translation
- conversation transcripts
Challenges and risks with AI
- Bias can affect the results
- Errors may cause harm
- Data could be exposed
- Solutions may not work for everyone
- Users must trust a complex system
- Who’s liable for AI-driven decisions?
6 Principles of Microsoft Responsible AI
- Fairness
- Reliability and Safety
- Privacy and Security
- Inclusiveness
- Transparency
- Accountability
Fairness Examples
- no data bias (gender in loan-approval)
- bias could be based on gender, ethnicity, or other factors that result in an unfair advantage or disadvantage to specific groups of applicants
- no facial recognition of emotions or identity attributes
Reliability and Safety Examples
- autonomous vehicles
- rigorous testing and deployment management
- patient diagnosis and recommendations for prescriptions
- can result in substantial risk to human life
- Handling of unusual or missing values provided to an AI system
- Ensure that AI systems operate as they were initially designed, respond to unanticipated conditions, and resist harmful manipulation
Privacy and Security Examples
- data origin and lineage
- data use internal vs. external
- data corruption considerations
- anomaly detection
- medical diagnostic bot
- Provide consumers with information and controls over their data collection, use, and storage
- Personal data must be visible only to approve
Inclusiveness Examples
- no audio output for visually impaired
- should benefit all parts of society, regardless of physical ability, gender, sexual orientation, ethnicity, or other factors
- designing an AI system that empowers everyone, including people who have hearing, visual, and other impairments
Transparency Examples
- Trust
- investment recommendations
- understanding purpose, how it works and limitations
-Design an AI system to assess whether loans should be approved, the factors used to make the decision should be explainable - Provide documentation to help developers debug code
- Automated decision-making processes must be recorded so approved users can identify why a decision was made
Accountability Examples
- innocent convicted based on facial recognition
- framework of governance and organizational principles
- ensure ethical and legal standards
- designers and developers
- Implementing processes to ensure that humans can override decisions made by AI systems
