Getting Started with AI

Artificial intelligence is no longer a distant promise. It writes code, composes music, diagnoses diseases, and holds conversations that are hard to distinguish from human ones. Whether you're a developer, a student, or just a curious reader, this guide will give you a solid foundation.

What is Artificial Intelligence?#

AI refers to computer systems that perform tasks that would normally require human intelligence — things like recognising speech, understanding language, making decisions, and learning from experience.

The field is broad. At the top level, it breaks into a few major subfields:

Subfield	What it studies	Example
Machine Learning	Systems that learn from data	Spam filters
Natural Language Processing	Understanding and generating text	ChatGPT
Computer Vision	Interpreting images and video	Face unlock on your phone
Robotics	Physical agents that sense and act	Warehouse robots

You don't need a maths PhD to get started with AI. Tools like Python, Jupyter notebooks, and libraries likescikit-learn mean you can run real models within minutes.

A Brief History#

AI has a surprisingly long history, punctuated by two "AI winters" — periods of dashed expectations and funding droughts.

1950 — Alan Turing proposes the "Turing Test"
1956 — The Dartmouth Conference coins "artificial intelligence"
1980s — Expert systems boom, then bust (AI winter #1)
1990s — Neural networks resurge, but slowly
2012 — AlexNet wins ImageNet, deep learning era begins
2017 — "Attention Is All You Need" — the Transformer paper
2022 — ChatGPT reaches 100M users in 60 days

How Machine Learning Actually Works#

Traditional programming is explicit: you write rules, and the computer follows them. ML flips this: you give the system examples, and it infers its own rules.

# Traditional approach
def is_spam(email: str) -> bool:
    return "buy now" in email.lower() or "limited offer" in email.lower()

# ML approach (conceptual)
from sklearn.naive_bayes import MultinomialNB

model = MultinomialNB()
model.fit(training_emails, training_labels)  # Learn from 10,000 examples

prediction = model.predict([new_email])  # Generalise to new data

The ML model figures out its own rules by finding patterns in the training data.

Key Concepts You'll Encounter#

Features and Labels#

• Features are the input variables the model sees (e.g., word counts in an email)
• Labels are the correct answers you're trying to predict (e.g., spam or not spam)

Training vs Inference#

• Training is the learning phase — the model adjusts itself by looking at labelled examples
• Inference is using the trained model to make predictions on new, unseen data

Overfitting#

A model that memorises its training data rather than learning general patterns will fail badly on new data. This is calledoverfitting. Always evaluate your model on data it hasn't seen.

Your First Step#

The best way to learn AI is by doing. Here's a minimal starting path:

1. Learn Python — the lingua franca of AI/ML
2. Get comfortable with NumPy and pandas — data manipulation fundamentals
3. Run your first model with scikit-learn (see our ML course)
4. Experiment with OpenAI's API to see LLMs in action

Google Colab gives you free GPU access and a Jupyter notebook environment with zero setup. It's the fastest way to run your first model.

What's Next?#

Now that you have the landscape in view, explore the rest of this blog. We have deep dives on transformer architecture, prompt engineering, and a structured course on machine learning fundamentals. Dive in wherever you feel most curious.