Frank Rosenblatt
(1928-1971)
Frank Rosenblatt studied psychology at Cornell, but for his Ph.D he built a computing machine, the Electronic Profile Analyzing Computer (EPAC), to perform multidimensional analysis for psychometrics.
He then joined the Cornell Aeronautical Laboratory in Buffalo, New York, where he began work on Perceptrons. He designed and built a computer that could learn new skills by trial and error. He called it the
Mark-1 Perceptron.
The Perceptron is a type of neural network which Rosenblatt believed could simulate human thought processes, based on the 1943 ideas of
Walter Pitts and
Warren McCulloch, who saw brain neurons as networks and the synapses as logic gates.
In the summer of 1958, Rosenblatt wrote an article for Cornell's publication
Research Trends. Click to read
The Design of an Intelligent Automaton
Four years later, he published
Principles of Neurodynamics: Perceptrons and the Theory of Brain Mechanisms..
Five years after Rosenblatt's book, in 1969, Marvin Minsky and Seymour Papert published their book
Perceptrons, in which they criticized Rosenblatt's ideas and doubted their applicability to complex problems.
From Perceptron to Transformer
Rosenblatt did not live to see the emergence of generative artificial intelligence, dying in a boating accident on his 43rd birthday in 1971. Nevertheless, Rosenblatt's perceptron, when stacked in layers as he imagined in 1958, is remarkably similar to the layers of
transformers in current AGI systems like Chat-GPT.
Normal |
Teacher |
Scholar
|
