AI 101: A beginner's guide to artificial intelligence

Demystify the "magic" of AI in plain English. From narrow vs general AI to LLMs, hallucinations, and how to talk to AI like an expert assistant.

Introduction: Demystifying the "magic"

For many, AI feels like a scene from a sci-fi movie — a sentient computer that "thinks" and "feels." In reality, AI is less about magic and more about extraordinary mathematics applied at lightning speed.

We are living through a shift similar to the invention of the steam engine or the internet. In the past, humans had to tell computers exactly what to do using rigid code. Today, we are teaching computers how to learn the rules for themselves. This tutorial is your roadmap to understanding this transition, the technology behind it, and how you can harness it without being a data scientist.

What you'll learn

Section 1 — The core definition: moving beyond the "robot" myth

Section 2 — The mechanics of learning: how data becomes intelligence

Section 3 — The generative revolution: why the world changed in 2022

Section 4 — Inside the LLM "brain": how AI understands human language

Section 5 — AI in the wild: hidden AI in your professional and personal life

Section 6 — The "ghost in the machine": hallucinations and bias

Section 7 — Mastering the interface: the fundamental logic of prompt engineering

Section 8 — The path forward: from consumer to builder

Section 1: What is AI?

At its simplest, Artificial Intelligence (AI) is a field of computer science that builds systems capable of performing tasks that typically require human cognition — such as recognizing faces, translating languages, or making complex decisions.

Diagram comparing Narrow AI as a specialized robot versus General AI as a versatile robot

The two flavors of AI

Narrow AI (ANI): This is what we use today. It is "narrow" because it is a specialist. An AI that can beat the world champion at Chess cannot tell you how to bake a cake. It excels in a predefined environment. Your Netflix recommendations, Siri, and face-recognition locks are all Narrow AI.

General AI (AGI): This is the "Holy Grail." An AGI would have the ability to learn any intellectual task that a human can. It could write a legal brief in the morning and repair a car engine in the afternoon. Currently, AGI exists only in research labs and science fiction.

Section 2: How AI "learns"

To understand AI, you must understand Machine Learning (ML). In traditional programming, a human writes a "recipe" in code: if the user clicks this, then do that. In Machine Learning, the computer creates its own recipe by looking at examples.

Pipeline diagram showing Data flowing into Algorithm flowing into Model

The three pillars of learning

1. The data (the experience) — AI needs "food." If you want an AI to identify a fraudulent transaction, you feed it millions of examples of both honest and dishonest bank records.
2. The algorithm (the teacher) — This is the mathematical framework. It looks at the data and tries to find correlations (e.g. most fraud happens at 3 AM from overseas IP addresses).
3. The model (the result) — Once the algorithm has finished looking at the data, the resulting "knowledge" is saved as a Model. You can then give this model new, unseen data, and it will make a prediction.

Section 3: Discriminative vs Generative AI

For decades, AI was mostly discriminative. It was a "judge." It looked at data and classified it: this is a cat, this is a spam email, this customer is likely to cancel their subscription.

The massive hype we see today is due to Generative AI.

Illustration of a glowing AI core radiating creative outputs like a paintbrush, code symbol, music note and text bubble

Generative AI doesn't just judge — it creates. Because it has analyzed the underlying structure of millions of images, poems, or lines of code, it can generate entirely new content that follows those same patterns.

When you ask a Generative AI for a picture of "A neon-lit Hong Kong in the year 2099," it isn't searching Google Images. It is "dreaming" up a new image based on its mathematical understanding of "neon," "Hong Kong," and "futuristic."

Section 4: Inside the LLM "brain"

When you interact with a tool like Gemini or ChatGPT, you are talking to a Large Language Model (LLM). This is a specific type of Generative AI trained on almost the entire public internet.

Chat input showing The capital of France is with Paris highlighted at 99 percent confidence

How it "speaks"

An LLM doesn't actually understand the meaning of words the way you do. Instead, it treats language as a statistical game of "what word comes next?"

If I say "The capital of France is...", your brain knows the answer is Paris. An LLM knows the answer is Paris because, in its trillions of pages of training data, the word "Paris" follows that specific sequence of words 99.9% of the time.

By scaling this up to billions of parameters, the AI can handle complex reasoning, summarize long documents, and even write software code. It is essentially a super-autocomplete.

Section 5: AI in the wild

You don't need to go to a special website to find AI — it is already the invisible "glue" of the modern world.

Smartphone surrounded by icons for navigation, shopping, healthcare and creative tools

Logistics & navigation — Apps like Google Maps use AI to analyze historical traffic patterns and real-time sensor data to predict traffic jams before they happen.

Personalized commerce — Amazon and Taobao use AI "collaborative filtering" to suggest products. If users A and B bought the same three items, and user A buys a fourth, the AI predicts user B will want it too.

Health tech — AI models are now better than some doctors at spotting early-stage cancer in X-rays because they can see pixel-level patterns invisible to the human eye.

Content creation — Tools like Cursor (for coding) or Canva (for design) use AI to handle the "boring" parts of creation, allowing humans to focus on the "vision."

Section 6: The "ghost in the machine"

As a beginner, it is vital to know that AI is confidently fallible.

1. Hallucinations

Because LLMs are predicting the next word based on probability, they can sometimes "hallucinate." They will state a fact that sounds perfectly logical but is completely made up. Never treat an AI as a source of truth without verifying facts.

2. Bias and "garbage in, garbage out"

If an AI is trained on data that contains human prejudice (like old hiring records that favored men), the AI will learn and amplify that bias. It doesn't know what is "fair" — it only knows what is in the data.

3. The black box

Even the engineers who build these models don't always know exactly why an AI made a specific decision. This is known as the "black box" problem, which is why AI is currently used as an assistant, not a final decision-maker.

Section 7: Mastering the interface

The most important skill in the AI era isn't coding — it's communication. The instructions you give an AI are called prompts.

Person at a desk sending a detailed instruction message to a friendly robot assistant

The "expert assistant" mental model

Instead of talking to the AI like a Google search bar, talk to it like a very smart, very fast, but literal-minded intern.

Be specific — Instead of "Write an email," say "Write a professional follow-up email to a client I met at a networking event yesterday."

Give examples (few-shot prompting) — If you want the AI to write in your style, paste three examples of your writing first and say "Analyze my tone and write the next paragraph in this style."

Iterate — Your first prompt is rarely the best. If the output is too long, tell it: "Good, but make it 50% shorter and use bullet points."

Section 8: The path forward

We are moving from a world where we use computers to a world where we collaborate with them. For a beginner, the goal is to develop AI intuition.

Don't fear the tech — You don't need a math degree. You just need curiosity.

Focus on the "why" — Let AI handle the how (the writing, the formatting, the coding), while you focus on the why (the strategy, the problem-solving, the human connection).

Start building — The best way to learn is by doing. Use AI to build a simple website, automate your emails, or summarize a book you've been meaning to read.