AI
Artificial Intelligence — the outer circle
The broadest possible category. Any technique at all that lets a machine mimic human intelligence counts as AI — rule-based chess engines from the 1960s, a thermostat that predicts your habits, and ChatGPT are all technically "AI."
Example: a simple if/else spam filter with hand-written rules is AI, even though it has no learning in it at all.
ML
Machine Learning — the middle circle
A subset of AI where the machine learns patterns from data instead of being explicitly programmed with rules. Every ML system is AI, but not every AI system is ML.
Example: a spam filter that learns from 10,000 labeled emails which words correlate with spam is ML — it built its own rules from data.
DL
Deep Learning — the innermost circle
A subset of ML that uses multi-layered neural networks (many stacked layers) to learn especially complex patterns. Every deep learning system is ML, and every ML system is AI — but the reverse is not true.
Example: an image classifier with 50 stacked layers recognizing cats vs. dogs is deep learning — the depth of layers is what makes it "deep."
Applied Walkthrough
1
A test question describes a system that follows a fixed decision tree with rules a human wrote by hand, no data-driven learning at all.
2
You ask: does it learn from data? No — it just follows pre-written logic.
3
Because it has zero learning component, it can only be AI. It is NOT ML, and definitely not deep learning.
4
Rule of thumb for exams: work from the inside out. If a system is deep learning, it is automatically also ML and also AI. If a system is only rule-based with no data-driven learning, it can only be AI.
Exam Application
Exam questions love to test this hierarchy directly: "Which of the following is true about AI and ML?" The correct answer is almost always the nesting relationship — ML is a subset of AI, and deep learning is a subset of ML. Any answer choice that treats these three as separate, parallel categories is wrong.
⚠ Common Trap
The single most common trap: assuming AI and ML are interchangeable synonyms, or assuming deep learning is a totally separate technique rather than a specific TYPE of machine learning. Every deep learning model IS a machine learning model — it just uses layered neural networks specifically. Also watch for questions that describe something clearly rule-based (no data, no learning) and mislabel it "machine learning" — rule-based systems without data-driven learning are AI only, not ML.
✓ Quick Self-Check
1. Is every machine learning system also an AI system?
Yes. ML is a subset of AI, so anything that qualifies as ML automatically qualifies as AI too.
Tap to reveal / hide
2. Is every AI system also a machine learning system?
No. AI is the broadest category and includes many non-learning systems (rule-based logic, search algorithms) that are not ML.
Tap to reveal / hide
3. What makes a neural network "deep" rather than just a regular neural network?
Having multiple stacked layers (often many) rather than just one or two. The "depth" refers to the number of layers.
Tap to reveal / hide
4. A chess engine from 1990 that only follows hand-coded rules with no learning — is it AI, ML, or DL?
AI only. With zero data-driven learning, it cannot be ML or DL, but the rule-based logic still counts as AI.
Tap to reveal / hide
5. True or false: ML and DL are separate, unrelated branches of AI.
False. DL is a subset of ML, not a separate branch — they are nested, not parallel.
Tap to reveal / hide