AI paradigms and their role in AI history

Part II, AI 100, 2026W1

Learning outcomes

Students will be able to:

  1. Compare and contrast explicit programming, rule-based systems, search, machine learning, deep learning, reinforcement learning, and self-supervised learning.
  2. Trace the historical development of AI paradigms and explain their significance in AI history.
  3. Implement or simulate small-scale examples of each paradigm.
  4. Explain the concepts of loss functions, gradient descent, and backpropagation.

Theme: Understand the key AI paradigms and their historical evolution to appreciate how AI techniques have developed over time.

Select topics

  • Explicit programming; rule-based systems.
  • Search: teach a computer to represent a problem and recognize a solution via algorithms and pattern matching.
  • Problem decomposition: agents, human-in-the-loop, graphical models.
  • Machine learning: infer patterns from examples, mapping features to labels
  • Deep learning: featureless inference, escaping the curse of dimensionality.
  • Reinforcement learning: learn complex policies based on sparse rewards.
  • Self-supervision and generative AI: eliminate dependence on manual labels by turning input data into completion puzzles (examples: next token prediction; image diffusion)
  • Key elements of neural network optimization: Loss functions, gradient descent, and back-propagation.