Era 06 / 15 · First AI Winter 1974–1980

The First Winter

Promises outran results, and the funding froze.

Beat 1 · Concrete

The frontier explodes

Each step multiplies the branches; the search frontier outruns every machine of the era.

Search tree exploding past the screen Each level of the tree triples the number of branches; the deepest level overflows the frame. ×3 branches every level → off the screen

Beat 2 · Abstract

Promise outran reality

Expectation soared while results stayed flat; in 1973 a report cut the line, and funding froze.

Promise versus reality, cut in 1973 Promise rises far above flat reality; the gap is coral; in 1973 the line plunges and color goes cold. 1973 · LIGHTHILL PROMISE REALITY

Beat 3 · Interactive

Push the problem size

Drag the slider: a few steps further and the required search blasts past any feasible machine.

Required search explodes past the feasible ceiling Bars stay tiny then blast off the top of the chart once the problem size crosses the feasible line. 1970s computer ≈ 1M nodes INTRACTABLE ↑
Why the cold set in

1973

The Lighthill Report

A UK government review judged AI a failure on its grand promises; "combinatorial explosion" was named as the wall. British funding was gutted overnight.

1969–1974

DARPA / SUR cuts

After the Mansfield Amendment and a disappointing Speech Understanding Research program, US defense money for open-ended AI dried up. Patience had run out.

The wall

Intractability

Toy problems scaled by brute-force search. Real ones branch exponentially: each extra step multiplies the work, so the frontier outgrows any computer long before the answer.