The Navy Promised a Thinking Machine in 1958. We Buried It for 17 Years — It's Inside AI Now.

On 7 July 1958, a psychologist named Frank Rosenblatt switched on a machine for the US Navy and the press. It had never seen the cards it was shown, it guessed at first, and after about fifty tries it started getting them right — with nobody writing the rule. The next day the New York Times reported that the Navy expected the "embryo of an electronic computer" that would "walk, talk, see, write, reproduce itself and be conscious of its existence." It was called the Perceptron. Eleven years later, a book by Marvin Minsky and Seymour Papert proved the simple version couldn't solve a four-case problem a child can do in a second — exclusive or. The proof was correct. The conclusion the field drew from it was not. Funding collapsed, Rosenblatt died in a boating accident on his 43rd birthday, and the idea sat in a drawer for seventeen years until backpropagation arrived in 1986 and taught the layer in the middle. This essay is about what that cost, and about what is actually inside the AI you use today: the same weighted sum, the same threshold, the same nudge toward right — scaled by a factor of about a billion. Rosenblatt wasn't wrong. He was early. And the story catches both of our favourite mistakes, hype and backlash, inside a single lifetime. Watch next (the series): ELIZA, the First Chatbot, Understood Nothing. People Told It Their Secrets Anyway.:    • ELIZA, the First Chatbot, Understood Nothi...   Deep Blue Didn't Outthink Kasparov — a Bug Did. And the Same Trap Is Inside AI.:    • Deep Blue Didn't Outthink Kasparov — a Bug...   The Horse That Fooled Scientists — and the Same Trap Hiding Inside AI (Clever Hans):    • The Horse That Fooled Scientists — and the...   The 250-Year-Old Trick Hiding Inside Modern AI (the Mechanical Turk):    • The 250-Year-Old Trick Hiding Inside Moder...   Tools this channel runs on: ElevenLabs (AI voiceover) → https://try.elevenlabs.io/x15ixv9ll3vd Chapters: 0:00 The machine the Navy said would walk, talk and see 1:38 Frank Rosenblatt, and learning from being wrong 3:24 The Mark I: 400 sensors and motorised dials 4:58 1969: Minsky, Papert, and exclusive or 6:41 The fix everyone knew, and nobody could train 8:15 1986: backpropagation, seventeen years late 9:45 The second winter — and the contest in 2012 11:16 What is literally inside a modern AI model 12:43 The hype cost money. The backlash cost a generation. Subscribe for weekly, no-hype AI breakdowns — hands on the tool, eyebrow raised. Some links above are affiliate links — signing up through them supports the channel at no extra cost to you. Music: "Frozen Star" Kevin MacLeod (incompetech.com) Licensed under Creative Commons: By Attribution 4.0 License http://creativecommons.org/licenses/b... — RESOURCES — The AI Revenue Lab Playbook: https://airevlab.gumroad.com/l/nvbyx Free 8-Niche Tier List: https://airevlab.gumroad.com/l/niches