AI vs. ML vs. Deep Learning
Artificial intelligence, machine learning, deep learning and generative AI are often used as if they mean the same thing. They don't: each is a smaller circle inside the previous one. Getting the terms right helps you read job adverts, news stories and research with a clear head.
The nested circles
- Artificial intelligence (AI) is the broad goal of building machines that perform tasks we associate with human intelligence: reasoning, understanding language, recognising images, planning.
- Machine learning (ML) is one way to achieve AI: systems that learn patterns from data rather than being given every rule.
- Deep learning (DL) is a type of machine learning that uses neural networks with many layers. It shines on images, audio and text.
- Generative AI is a use of deep learning where models create new content (text, images, code, audio), such as chatbots built on large language models.
So every deep learning system is machine learning, and every machine learning system is AI, but not the other way round.
AI without machine learning
Early AI was mostly rule-based: experts wrote long lists of if-then rules. A chess program that searches possible moves, or a route planner using a shortest-path algorithm, is AI with no learning involved. Here is a tiny rule-based "AI" for flagging suspicious mobile money messages:
SUSPICIOUS = ["send back", "wrong number", "pin", "you have won"]
def looks_like_scam(message):
text = message.lower()
return any(phrase in text for phrase in SUSPICIOUS)
print(looks_like_scam("I sent money to the wrong number, please send back"))
print(looks_like_scam("Meeting moved to 3 pm"))
This works until scammers change their wording. A machine learning model trained on thousands of labelled messages can pick up patterns no one thought to write down, and can be retrained as scams evolve.
Classic ML vs. deep learning
| Classic machine learning | Deep learning | |
|---|---|---|
| Best data | Tables (rows and columns) | Images, audio, text, video |
| Features | Often designed by people (feature engineering) | Learned automatically from raw data |
| Data needed | Hundreds to thousands of rows can be enough | Usually much more |
| Hardware | A normal laptop | Often GPUs for training |
| Explainability | Often easier (e.g. decision trees) | Harder; a "black box" |
| Examples | Loan scoring, sales forecasting | Face unlock, speech recognition, chatbots |
For tabular business data (sales, loans, survey results), classic ML such as random forests is often as good as or better than deep learning, and far cheaper. Don't reach for a neural network by default.
A short history
- 1956: the term "artificial intelligence" is adopted at a summer workshop at Dartmouth College in the USA.
- 1959: Arthur Samuel popularises the term "machine learning" while building a checkers program that improved by playing.
- 1980s–2000s: statistical methods such as decision trees and support vector machines become practical.
- 2012: a deep convolutional network (AlexNet) wins the ImageNet image-recognition challenge by a wide margin, starting the deep learning boom.
- 2017: researchers at Google introduce the transformer architecture in the paper "Attention Is All You Need".
- 2022 onwards: chatbots built on large language models, such as ChatGPT (released November 2022), bring generative AI to the public.
Narrow AI vs. general AI
Every AI system in use today is narrow: it is built and trained for particular kinds of tasks. Even capable chatbots can make basic mistakes outside their training. Artificial general intelligence (AGI), a system matching humans across all intellectual tasks, is a research goal and a subject of debate, not an existing product.
Calling every automated system "AI" can hide how it really works. When someone says a product "uses AI", ask: is it rules, classic ML, or deep learning? What data trained it, and how was it tested? Those answers tell you far more than the label.
Key points
- AI is the broad field; ML is AI that learns from data; deep learning is ML with many-layered neural networks.
- Generative AI uses deep learning to create new content.
- Classic ML suits tabular data; deep learning suits images, audio and text.
- All current AI is narrow; general AI does not yet exist.
Tip: use the ← and → keys to move between pages.
