The hybrid AI revolution

Most of the early AI systems were expert systems powered by rule-based system and inference engines, dominated AI from 1970s till 1990s.

- RAJENDRA M SONAR, PH.D., SJMSOM, IIT BOMBAY

An expert system is an AI program that emulates human expert decision-making in specialized domains using knowledge-based rules. These systems, which are highly effective for deterministic, complex, narrow, and heuristic-based problems, improve consistency and accessibility of expert knowledge. technology strategy.

Expert systems were developed in many domains and verticals since 1970s. MYCIN was a pioneering 1970s Stanford University rule-based expert system designed to identify bacteria causing severe infections and recommend antibiotics. Using 600 production rules and backward chaining (goal driven problem solving), it achieved accuracy comparable to infectious disease specialists. It was notable for explaining its reasoning, though never used clinically. There were many applications developed in BFSI domain in 1980s till 2000s. Let us start exploring on loan underwriting domain as a reference use-case and specifically CLUES (Countrywide Loan Underwriting Expert System). CLUES was a pioneering rule-based artificial intelligence system launched in February 1993 and was designed to automate the mortgage underwriting process, increasing loan production capacity while maintaining consistency and reducing cost. It paved the way for modern automated underwriting systems (AUS) like Fannie Mae's Desktop Underwriter (DU) and Freddie Mac's Loan Product Advisor (LPA). By the mid-1990s, the system was processing over 8,500 loans monthly across more than 300 decentralized branches and was handling 35% of all Countrywide loan volume. CLUES contained approximately 1,000 business rules derived from human underwriting expertise and secondary market guidelines from more than 100 hours of interviews were conducted.

What AI CTOs did in 1990s