"3. Neuron Nets: How can a set of (hypothetical) neurons be arranged so as to form concepts. Considerable theoretical and experimental work has been done on this problem by McCulloch and Pitts, Hebb, and Rochester."
ā McCarthy, Minsky, Rochester, & Shannon (Dartmouth Proposal, Aug 31, 1955)
Neuron Net: McCulloch-Pitts & Synaptic Threshold Logic
1955 Conjecture vs Modern 2020s AI Paradigm
1955 Historical Hypothesis
McCarthy & Minsky posited that trial-and-error reinforced threshold assemblies of artificial neurons could learn concept formation and abstract logic.
Modern Deep Learning Outcome
Realized at planetary scale: Transformer neural attention layers (GPT-4, Claude) with billions of continuous backprop synaptic weights replacing binary step functions.
Lineage of the 1955 Dartmouth Summer Project
1943-1951
Origins: McCulloch-Pitts formal neuron & Minsky's SNARC vacuum tube learner.
Aug 31, 1955
Dartmouth Proposal: McCarthy, Minsky, Rochester, & Shannon draft the charter.
Summer 1956
The Workshop: Newell, Simon & Shaw demonstrate the Logic Theorist in Hanover.
Modern Era
Scaling Laws: 7 original conjectures fulfilled via Transformer architectures & GPU compute.