"Knowledge is power: Modeling expertise through structured rules."
Dedicated hardware for AI development becomes a billion-dollar industry, allowing companies to run complex symbolic programs.
Japan invests heavily in massively parallel computing to leapfrog current AI capabilities, triggering a global AI arms race.
Digital Equipment Corp (DEC) deploys XCON to save millions by automatically configuring computer orders. This proved AI had massive ROI.
Rumelhart, Hinton, and Williams publish on **Backpropagation**, showing how multi-layer neural networks could finally be trained.
Douglas Lenat begins "Cyc," an ambitious attempt to codify "common sense" into a giant database of millions of assertions.
Unlike previous "general" AI, Expert Systems separated the "brain" from the "memory":
The processing unit that applies logical rules to the data to deduce new facts (Forward vs. Backward Chaining).
A repository of domain-specific facts and "if-then" rules gathered from human specialists.
Starts with known data and applies rules to see what results occur (Data-driven).
Starts with a goal (hypothesis) and looks back for data that supports it (Goal-driven).
While successful, these systems were static. They couldn't learn on their own. If a doctor's knowledge changed, the "Knowledge Engineer" had to manually rewrite thousands of lines of code. This led to the 2nd AI Winter when maintenance costs exceeded the benefits.