Andrew Ng AI for Everyone

Artificial Intelligence (AI)

The science of making computers perform tasks that typically require human intelligence — such as understanding language, recognizing images, and making decisions. e.g. spam filters, voice assistants, recommendation systems

Artificial Narrow Intelligence (ANI)

AI that is good at one specific task. It cannot generalise beyond its training. This is the only type of AI that exists today. e.g. Chess engines, face recognition, self-driving car perception

Artificial General Intelligence (AGI)

Hypothetical AI that can perform any intellectual task a human can — reasoning, learning, and adapting across domains. Does not yet exist. e.g. A machine that can write code, do surgery, and compose music equally well

Machine Learning (ML)

A branch of AI where systems learn patterns from data to make predictions or decisions — without being explicitly programmed with rules. e.g. Learning that certain emails are spam, from historical examples

Supervised Learning

The most common ML technique. You give the model labelled input-output pairs (A → B), and it learns to map new inputs to correct outputs. e.g. Input: house size → Output: price; Input: email → Output: spam/not spam

Unsupervised Learning

ML without labels. The algorithm finds hidden structure or clusters in data on its own — no "right answer" is provided. e.g. Customer segmentation, topic modelling in documents

Reinforcement Learning

An agent learns by taking actions in an environment and receiving rewards or penalties. It optimises for maximum cumulative reward over time. e.g. AlphaGo learning to play Go; a robot learning to walk

Deep Learning

A subset of ML using multi-layered neural networks. Extra layers let the model learn increasingly abstract representations of data. e.g. The engine behind modern image recognition, speech, and LLMs

Neural Network

A computational model loosely inspired by the brain. Layers of interconnected nodes (neurons) transform inputs into outputs through learned weights. e.g. Input layer → hidden layers → output layer predicting a house price

Features

The input variables fed into an ML model. Choosing good features (feature engineering) is a crucial part of building effective models. e.g. House size, number of bedrooms, and zip code are features for predicting price

Data Science

The process of extracting insights and value from data using statistics, analysis, and visualization. Output is often a report or dashboard rather than a deployed model. e.g. Analyzing sales trends to recommend which product lines to expand

Structured Data

Data organized in rows and columns — like a spreadsheet or database table. Easier for traditional ML algorithms to work with directly. e.g. A table of customers with columns: age, city, purchase amount

Unstructured Data

Data without a predefined format — images, audio, video, and raw text. Deep learning excels at extracting meaning from unstructured data. e.g. Medical X-ray images, call centre audio recordings, social media posts

Generative AI (GenAI)

AI that can create new content — text, images, code, audio — by learning patterns from large datasets. Powered by large language models and diffusion models. e.g. ChatGPT writing an email; DALL-E generating an illustration

Large Language Model (LLM)

A very large neural network trained on massive text corpora to predict and generate language. The foundation of most modern GenAI tools. e.g. GPT-4, Claude, Gemini — used for writing, coding, Q&A

Prompt Engineering

The practice of crafting inputs (prompts) to guide an LLM toward the desired output. A key skill for getting value from GenAI tools without retraining them. e.g. Adding "Think step by step" to a prompt can dramatically improve reasoning quality

Algorithm

A set of rules or steps a computer follows to solve a problem or learn from data. In ML, the algorithm defines how the model learns from examples. e.g. Linear regression, decision trees, and gradient descent are all algorithms

AI Project Workflow of ML Project

Andrew Ng's framework: (1) Collect data, (2) Train model, (3) Deploy & monitor. Real projects iterate through these steps repeatedly. e.g. Deploy a speech model → notice errors in noisy environments → collect more noisy audio → retrain

AI Project Workflow of DS Project

Andrew Ng's framework: (1) Collect data, (2) Analyze data, (3) Suggest Hypotheses/actions. Real projects iterate through these steps repeatedly. e.g. Deploy changes → re-analyze new data periodically

Model

The output of training an ML algorithm on data. It's the learned function that takes new inputs and produces predictions. e.g. After training on house data, the model can predict the price of a new listing