The verdict

In 1969, Marvin Minsky and Seymour Papert published a book called Perceptrons.

It proved, with mathematics, that the Perceptron had a hard limit. Not a flaw you could fix. A ceiling built into the design itself.

The proof was correct.

What happened next wasn't. The field took it as a verdict on the entire approach — not just this one design, but the whole idea of machines that learn from examples. Funding dried up almost overnight. Researchers moved on. The Perceptron became a cautionary tale.

Minsky wasn't a neutral observer here. He had been at Dartmouth in 1956, one of the people who believed AI should be built from rules and logic — the approach Rosenblatt was directly challenging. He had a reason to want the answer to go a certain way.

The proof was right. The conclusion the field drew from it was not.