Lecture Notes: AI and Data Formats

Historical Milestones and Future of AI

• Evolution of AI over the last 30 years.
• Predictions for AI advancements in the coming decades.

Data Formats in AI

• Structured Data: Spreadsheets, relational databases (e.g., msql, myql, Oracle).
• Semi-structured Data: JSON, XML, CSV (e.g., mongodb, couchDB).
• Unstructured Data: Text (PDFs), Images (JPEGs), Audio (MP3).

Distinguishing Data Science and AI

• Data Science: Applying machine learning to structured data.
• AI: Applying machine learning to unstructured data to simulate human intelligence tasks like speaking, reading, visual perception.

Example Scenarios

• Structured data: Used in enterprise data warehouses.
• Text data: Applications in text analysis.

Computer Vision

• Definition: Interpret, understand, and make decisions based on visual data.
• Common Applications:
  • Image Segmentation: Drawing boxes around objects.
  • Object Detection: Identifying specific objects (e.g., finding an ambulance in a drone image).
  • Facial Recognition: Used in devices for security.
  • Edge Detection: Identifying object boundaries.
  • Pattern Detection: Finding automatic patterns in data (e.g., satellite imagery).
  • Image Classification: Categorizing images into predefined categories.
  • Feature Matching: Comparing images based on visual features.
• Computer Vision Use Cases:
  • Object Detection and Recognition: Self-driving cars, traffic management, video surveillance, retail analytics, disaster response, environment monitoring, content retrieval.
  • Anomaly Detection: Identifying unusual events/objects (e.g., misplaced objects, abnormal behavior).
  • Face Recognition: Security, unlocking devices, law enforcement, identity verification, attendance tracking.
  • Video Summarization: Creating condensed, information-rich videos from longer recordings.

Audio Analytics

• Speech Recognition: Converting spoken language (audio) to text.
• Speaker Recognition: Identifying who is speaking based on audio signals.
• Sound Analysis: Identifying and classifying non-speech sounds (e.g., factory noises, environmental sounds).

Computational View of AI

• All data can ultimately be represented in a relational format (e.g., spreadsheet/table).
• Feature Engineering: Transforming unstructured data into a structured format for use in AI/ML models.

Key Types of Machine Learning

• Unsupervised Learning: Finding patterns without labeled outcomes.
  • Clustering: Grouping similar data points.
  • Anomaly Detection: Identifying unusual data points.
  • Recommender Systems: Predicting user preferences and suggesting items.
• Supervised Learning: Inferring a function to predict outcomes based on labeled data.
  • Regression: Predicting continuous outcomes (e.g., sales based on advertising spend).
  • Classification: Predicting categorical outcomes (e.g., spam detection in emails).