CONCEPTS & TURNING POINTS / 1 MIN READ
Expert Systems and the First AI Winter (1970s–1980s)
On this page
Expert systems brought AI to practical, narrow domains but also fueled overpromising that led to funding collapse. This sub-article expands Section 4 of History.
Expert Systems
- Rule-based systems encode human expertise as if–then rules.
- MYCIN (1975) diagnosed infections and recommended antibiotics; XCON configured computer systems.
- Knowledge representation — frames, semantic networks, and logic programming (Prolog) aimed to capture domain knowledge.
The Fifth Generation Project
- Japan’s Fifth Generation Computer Systems (1982) aimed at logic-based AI.
- Western programs (UK Alvey, US Strategic Computing) responded; the project ultimately delivered less than promised.
The First AI Winter
- Funding collapse: overpromising, underdelivery, and the LISP machine market crash (1987) reduced funding.
- Many commercial AI firms failed or pivoted.
- Lesson: hype cycles recur; capability must be separated from marketing.
Why It Matters
Expert systems proved AI could be economically useful, but also revealed that hand-coded knowledge does not scale. This tension set the stage for statistical learning.
Sources
- Buchanan, B. & Shortliffe, E. (1984). Rule-Based Expert Systems: The MYCIN Experiments.
- Feigenbaum, E. et al. (1988). The Fifth Generation Project.
Back to History · See also AI Winters · Next: Statistical Learning.