AI, ML and DL

You have heard all three of these. People use them as if they meant the same thing. They do not — and the exam is far more interested in how they fit together than in any one definition.

1The three terms

Artificial Intelligence (AI)

Any technique that lets a computer mimic human intelligence. An artificially intelligent machine works on the algorithms and the data fed to it, and gives the output you wanted.

Machine Learning (ML)

Machines that get better at a task with experience. The machine learns from the new data it is fed while testing, and uses that for the next round. It also takes account of the times it went wrong.

Deep Learning (DL)

Software that trains itself, using vast amounts of data. Because it has such a huge set of data, it can train itself with several machine learning algorithms working together.

2Which sits inside which

This is the part that is actually examined. They are not three things standing side by side — they are three circles, each inside the last.

Three words, one inside the next

Pick a term to light up its shell. Drag sideways to turn the stack and see how deep each one sits.

The dots are AI projects. There are a lot of them; a few of those are machine learning; very few of those go all the way into deep learning.

Read it outward: DL is inside ML, and ML is inside AI.

One sentence to remember

AI is the umbrella term. It holds Machine Learning and Deep Learning under it. Deep Learning is a subset of Machine Learning, because it is made of several machine learning algorithms working together.

Russian dolls

Open the AI doll and you find the ML doll inside it. Open that and the DL doll is inside that. You can never find the big doll inside the small one — and that is exactly the mistake the question is looking for.

The handbook calls it a funnel: there are a great many applications of AI, out of which a few come under ML, out of which very few go all the way into DL.

3ML and DL, side by side

Both learn from data. The difference is how much data, and how much help they need from the programmer.

	Machine Learning	Deep Learning
Learns from	Past or historical data fed to it	Vast amounts of data
Features	Often pointed out by the programmer	Extracted automatically, with no input from the programmer
Built from	One learning algorithm	Many ML algorithms working together, as a neural network
Examples	Object classification, anomaly detection	Object identification, handwritten digit recognition

Two ways to recommend a film

You liked Dangal and Sultan.
Machine Learning notices both are tagged “Aamir Khan” and “sports drama”, and suggests Bhaag Milkha Bhaag, which carries the same tags. Somebody had to write those tags first.
Deep Learning has no tags. It looks at the colours of the posters, the background music and the plot summaries of what you watched, works out your taste on its own, and suggests something you might love even though nothing obvious connects it.
That is automatic feature extraction: ML works from features a human wrote down, DL finds its own.

What those examples mean

Object classification decides which category a thing belongs to. Anomaly detection finds the unexpected thing hiding in the data — a sudden spike in a heart-rate reading, for instance. Digit recognition is training a computer to read handwritten digits 0–9 out of an image.

4From the board papers

Srishti learnt about AI terminologies but could not recall the term used for machines that perform tasks with vast amounts of data using neural networks. Help her with the correct term.CBSE sample paper 2024–25

1 mark

▶Show the answer

Deep Learning (DL). Two clues in the question point only at DL: vast amounts of data and neural networks. Machine Learning on its own does not require either.

Differentiate between AI, ML and DL. (Practice)

2 marks

▶Show the answer

AI is any technique that enables a computer to mimic human intelligence. ML is a subset of AI that enables machines to improve at a task with experience, learning from the data fed to it. DL is a subset of ML that enables software to train itself with vast amounts of data, using several machine learning algorithms together.
AI is the umbrella term: DL lies inside ML, and ML lies inside AI.

5Check yourself

Quick Check

Which of these is the umbrella term that holds the other two?

Quick Check

Deep Learning is a subset of:

Quick Check

Which one can extract features from data automatically, without the programmer pointing them out?

← Previous

📋 Bioethics: A Case Study

Data, Features and Labels 🏷️