Two Paths of AI

Module 2 · Section 1 of 3

In 1997, a computer beat the world chess champion. The headlines called it a milestone for machine intelligence. What they missed was the more important story: how that computer worked — and how different it was from the AI you use today.

Deep Blue didn’t think. It calculated. Millions of positions per second, evaluated against rules written by human experts. It was extraordinarily good at one specific thing and completely useless at everything else.

That version of AI hit a wall. The problems worth solving don’t come with complete rule sets. The world is too messy, too context-dependent, too large to pre-program your way through.

So AI went in a different direction. Instead of writing rules for machines to follow, researchers started letting machines discover their own rules by learning from data. The result was something neither approach fully predicted: systems that can hold a conversation, write code, summarise a document, and explain their reasoning — all in the same session.

The two paths — programmed rules versus learned patterns — are still both in use. Most AI tools you encounter today combine them. Understanding what each one does well, and where each one breaks, changes how you work with these tools.

That’s what this module covers.