Short answers, pulled from the story.
Frank Rosenblatt invented the perceptron in 1957 while working at the Cornell Aeronautical Laboratory. He first simulated it on an IBM 704, then built the physical Mark I Perceptron machine, which was publicly demonstrated on the 23rd of June 1960.
The Mark I Perceptron was designed for image recognition. It had three layers: a 400-photocell input retina arranged in a 20 by 20 grid, 512 association units in a hidden layer, and 8 response units in an output layer. It is now held at the Smithsonian National Museum of American History.
At a 1958 press conference organized by the US Navy, Frank Rosenblatt made claims about the perceptron's potential that The New York Times described as predicting a machine that could "walk, talk, see, write, reproduce itself and be conscious of its existence." These statements provoked sharp disagreement among AI researchers.
Minsky and Papert proved that single-layer perceptrons cannot learn the XOR function, a basic logical operation requiring a nonlinear decision boundary. Despite widespread misreading, the book did not claim multilayer networks faced the same limits; both authors already knew multilayer perceptrons could solve XOR.
The 1969 Perceptrons book caused a significant decline in interest and funding for neural network research. The field did not recover until the 1980s, roughly ten years later. A corrected expanded edition of the book was published in 1987.
Rosenblatt continued working on perceptrons despite shrinking funding. His final hardware project, Tobermory, was built between 1961 and 1967 for speech recognition, occupying an entire room with 12,000 weights. He died in a boating accident in 1971, before neural network research revived in the 1980s.