Lecture 07 — AI Fields & Specializations

~2–2.5 hours (AI landscape lecture)


🌐 Big Truth

AI is not one field.
AI is an ecosystem of specialized intelligences.

Understanding AI means understanding:

  • what problem you are solving
  • which field fits that problem
  • how fields connect

🧭 The AI Landscape (Mental Map)

Think of AI like a city 🏙️:

  • NLP = language district
  • Vision = visual district
  • RL = decision-making district
  • Generative AI = creative district
  • LLMs = central brain

Fields overlap and collaborate.


🗣️ 1️⃣ Natural Language Processing (NLP)

📘 What it is

Teaching machines to understand and generate human language.


🧠 Core problems

  • tokenization
  • syntax
  • semantics
  • context

🧪 Applications

  • translation
  • chatbots
  • search engines
  • summarization

💻 Mini Example

text = "AI changes the world"
tokens = text.split()

🎯 Mini Project Idea

Build a:

  • sentiment analyzer
  • document summarizer
  • chatbot

👁️ 2️⃣ Computer Vision (CV)

📘 What it is

Teaching machines to interpret images and videos.


🧠 Core problems

  • object detection
  • segmentation
  • tracking

🧪 Applications

  • self-driving cars
  • medical imaging
  • facial recognition

💻 Mini Example

image.shape  # height x width x channels

🎯 Mini Project Idea

  • face detector
  • image classifier
  • medical image analyzer

🧠 3️⃣ Large Language Models (LLMs)

📘 What they are

Very large neural networks trained on massive text data.


🧠 What makes them special

  • emergent reasoning
  • instruction following
  • tool usage
  • multimodal ability

🧪 Examples

  • ChatGPT
  • Gemini
  • Claude

🎯 Mini Project Idea

  • Q&A bot
  • code assistant
  • research summarizer

🎨 4️⃣ Generative AI

📘 What it is

AI that creates new content.


🧠 Models

  • Diffusion models
  • GANs
  • Autoregressive models

🧪 Applications

  • image generation
  • music composition
  • video synthesis

💻 Mini Concept

Noise → Model → Image

🎯 Mini Project Idea

  • image generator
  • style transfer
  • text-to-image demo

🕹️ 5️⃣ Reinforcement Learning (RL)

📘 What it is

Learning through interaction and reward.


🧠 Core idea

No labels. Only feedback.


🧪 Applications

  • game AI (AlphaGo)
  • robotics
  • recommendation systems

💻 Mini Concept

State → Action → Reward → Update

🎯 Mini Project Idea

  • game-playing agent
  • robot controller
  • decision optimizer

🔀 6️⃣ Multimodal AI (Where Everything Meets)

📘 What it is

AI that understands multiple data types at once.


🧪 Examples

  • GPT-4V
  • Gemini
  • CLIP

🧠 Why it matters

The real world is multimodal.


🧠 Fields Are Not Isolated

Example:

  • Self-driving car = CV + RL + sensors
  • ChatGPT = NLP + LLM + RLHF
  • Medical AI = Vision + NLP + statistics

🎓 Career Path Mapping (Very Useful)

Interest Field
Language & logic NLP / LLM
Images & perception CV
Decision-making RL
Creativity Generative AI
Systems thinking Multimodal AI

🌱 Final Big Insight

Great AI systems are built by combining fields — not mastering only one.

Understanding the map makes you powerful.


❓ Final Quiz

Can one model belong to multiple fields?

Yes — modern AI systems are deeply interdisciplinary.

Previous
Next