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.