AI and Machine Learning

When studying for an exam, you can either memorize a cheat sheet word for word or solve thousands of practice questions until you get the hang of it. In computer programming, the exact same two approaches exist. Traditional programming means a human writes down strict rules for a machine to follow. Machine learning, on the other hand, is letting the computer figure out the rules and intuition on its own by studying data. The most common method shows the computer countless questions paired with correct answers so it discovers the underlying formula—a method known as supervised learning. Machine learning also includes finding hidden patterns without answer keys or learning through trial, error, and rewards.

All of this falls under one grand mission: artificial intelligence (AI). Rather than a single specific technology, AI is the broadest umbrella and ultimate goal—creating machines that mimic human intelligence.

Most modern AI we meet every day, such as voice assistants on our phones, language translators, and recommendation feeds, is built using this machine learning study method.

Tasks that feel effortless to humans can be brutally difficult for computers. Imagine writing explicit code to tell a computer whether a photo shows a cat or a dog. If you define a cat as 'an animal with pointy ears, fur, and whiskers,' the computer gets completely lost when it encounters a Scottish Fold with folded ears or a fluffy dog.

Humans tell them apart instantly, but translating that gut instinct into words and code is nearly impossible. Instead of spelling out every rule, machine learning hands the computer tens of thousands of labeled photos.

By comparing countless images, the computer uncovers statistical patterns on its own: 'Aha, cats usually have these eye shapes and contours.' This allows the machine to pick up on subtle nuances that humans could never fully explain.

In the news, AI, machine learning, and deep learning are often tossed around interchangeably, but they actually nest inside one another like Russian Matryoshka dolls. The biggest, outermost doll is artificial intelligence, the overarching vision of any machine that exhibits human-like intelligence.

The medium doll nested inside is machine learning, which focuses on letting computers learn directly from data. Decades ago, a chess program driven solely by hardcoded rules was considered AI, but it was not machine learning because it never learned from data.

Deep inside machine learning sits deep learning. Inspired by the neural networks of the human brain, it stacks multiple layers of artificial neurons to process complex data like photos, audio, and text. While the concept existed for decades, it exploded in the 2010s thanks to massive datasets and powerful computing chips. In short, deep learning is a branch of machine learning, and machine learning lives inside the broader world of AI.