What Is Artificial Intelligence?

A chess engine, a spam filter, and a chatbot all get called "AI," but they work nothing alike under the hood. Before this course goes any further, it's worth pinning down what the term actually covers.

A working definition

Artificial intelligence, broadly, is the effort to build systems that perform tasks which would normally require human intelligence — recognizing a face, translating a sentence, deciding the best next move in a game, predicting whether a transaction is fraudulent. Notice this definition says nothing about how the system does it. A chess engine that brute-force searches millions of positions and a language model that predicts text one token at a time are both "AI" by this definition, even though they share almost no machinery.

Narrow AI vs. general AI

Every AI system in real-world use today — including the most impressive ones — is narrow AI: built for a specific task or class of tasks, with no ability to transfer that competence somewhere else. A system that plays chess at a superhuman level cannot fold laundry, and a system that writes fluent paragraphs cannot, on its own, drive a car. General AI (sometimes called AGI) — a single system with human-level competence across essentially any task — does not exist yet. It's a research goal and a subject of real disagreement among experts about whether, or when, it's achievable, not something already running in a product you can buy.

Common misconceptions worth clearing up early

  • "AI" doesn't imply consciousness, feelings, or genuine understanding — a system can produce very convincing output while doing nothing resembling what a human means by "thinking."
  • "AI" isn't one technology. It's an umbrella covering search algorithms, statistical machine learning, neural networks, and more — this course covers several of them separately because they really are different tools.
  • Impressive output doesn't imply general capability. A model that writes excellent code can be confidently, fluently wrong about basic arithmetic.
The "AI effect": there's a long-observed pattern where, once an AI technique becomes well understood and reliable, people stop calling it "AI" and start calling it just "software." Optical character recognition and chess engines were both cutting-edge AI research in their day; today they're considered ordinary, solved problems. Keep that in mind whenever "AI" is used as if it names one fixed, stable thing — the boundary of the term keeps moving.