Chapter 1 · The dawn of AI (1943–1990)1980s
Expert systems
Expert systems packed specialist knowledge into thousands of rules; powerful but brittle and costly.
Chapter 1 · The dawn of AI (1943–1990)1980s
Expert systems packed specialist knowledge into thousands of rules; powerful but brittle and costly.
The rule book matches symptoms to a fault and explains why. Then an unusual car arrives that no rule describes, and the system has nothing to say.
In the 1970s and 80s, builders called knowledge engineers interviewed doctors, engineers and other specialists and wrote what they said as IF-THEN rules. An early example, MYCIN at Stanford, suggested treatments for blood infections.
Tick symptoms, or tap a test case, and read the reasoning. Then try the cases the rule book misses and write a rule for them. A tiny teaching example with made-up confidence numbers.
Add a few rules and watch the coverage creep up. With only seven symptoms, hand-writing a rule for every combination quickly becomes impossible.
Where did an expert system's knowledge come from?
What does brittle mean for an expert system?
Why was it costly to keep an expert system useful?