Artificial Intelligence (AI) is the field of building systems that perform tasks which normally require human intelligence — understanding language, recognizing images, making decisions, playing games. Deep Learning, the subject of this entire notes hub, is one specific technique inside AI. Before going deep into neural networks, it helps to see the whole map first.
What "Intelligence" Means for a Machine
A system counts as AI if it can take in information, reason or pattern-match over it, and produce a useful action or prediction — without a human manually scripting every possible situation in advance. That's a deliberately broad definition, because AI is an umbrella covering many very different techniques: hand-written logic rules, search algorithms, statistical machine learning, and deep neural networks are all "AI."
The Sub-Fields Under the AI Umbrella
| Sub-field | What It Studies | Typical Technique Today |
|---|---|---|
| Machine Learning (ML) | Learning patterns from data instead of hand-coded rules | Decision trees, gradient boosting, neural networks |
| Deep Learning (DL) | Learning with multi-layer neural networks | CNNs, RNNs, Transformers |
| Natural Language Processing (NLP) | Understanding and generating human language | Transformer-based language models |
| Computer Vision (CV) | Understanding images and video | CNNs, Vision Transformers |
| Robotics | Perceiving and acting in the physical world | Reinforcement learning + control theory |
| Expert Systems (classical AI) | Encoding human expert knowledge as rules | If-then rule engines (largely historical) |
Deep Learning is not a separate field sitting next to ML — it is inside ML, and it is the main engine currently powering modern NLP and Computer Vision. The next note, AI vs ML vs Deep Learning, draws that containment relationship precisely.
Narrow AI vs General AI vs Super AI
| Category | Definition | Example | Status Today |
|---|---|---|---|
| Narrow AI (ANI) | Excels at one specific task only | Image classifier, spam filter, chess engine, GPT-style chatbot | This is everything that exists today, including every model this course covers. |
| General AI (AGI) | Human-level reasoning across any task | A system that can learn any intellectual task a human can | Does not exist yet — an active research goal. |
| Super AI (ASI) | Intelligence exceeding humans across all domains | Hypothetical | Speculative; not a near-term engineering target. |
Every deep learning model you will build in this course — an image classifier, a chatbot, a translator — is narrow AI. It is extremely good at the one thing it was trained for and has no general reasoning ability outside that.
Why Start a Deep Learning Course Here
Framing matters for how you evaluate every technique that follows. When you later ask "why use a CNN instead of a plain neural network for images?" or "why use a Transformer instead of an RNN for text?", the honest answer is always the same shape: a more specialized architecture encodes assumptions about the data that a generic one doesn't, so it learns the task with less data and compute. AI → ML → DL is the same story at a larger scale: each narrower field trades generality for effectiveness on its target problems.
Common Mistakes
- Treating "AI" as synonymous with "deep learning" — AI is the umbrella; deep learning is one technique under it (see the next note for the exact relationship).
- Assuming today's narrow AI systems are close to general intelligence because they sound fluent — a large language model has no world model or goals; it produces the statistically likely next token.
Interview Relevance
Q: "Is ChatGPT an example of AGI?" No — it is a very capable narrow AI system, extremely good at text generation, with no general reasoning or autonomous goal-pursuit outside that. Being fluent in language is not the same as general intelligence.
Practice Question
List three AI systems you use in a normal week (e.g. a recommendation feed, a voice assistant, a spam filter). For each, name which AI sub-field(s) it most likely relies on.
