The Promise and the Crash

Confidence is easiest at the beginning.

When a new idea first starts working, it's natural to assume the rest is just a matter of time and effort. The hard part, it seems, is proving the thing is possible at all. Once that's done, surely the details follow.

This is a pattern that repeats throughout the history of AI. Early results arrive and they're genuinely impressive. Progress feels fast. Predictions get made. Timelines get announced. And then the problems that seemed like engineering challenges turn out to be something deeper.

Language requires the entire world as context. Reasoning requires knowledge, and knowledge turns out to be effectively infinite. Every solution opens onto a larger problem. The cliff that looked short from the bottom is, in fact, very tall.

The first AI winter was not a story of bad science. It was a story of early science, misread as proof of something it hadn't yet proved. A foothold mistaken for a summit.

The gap between what was promised and what was possible is the story of this chapter.