The grand challenge facing the pioneers of artificial intelligence (creating an ‘intelligent’ computer) was simply this: if a computer could think like a human how could we tell? This famous test devised by Alan Turing has, arguably, yet to be passed by a computer after over fifty years ... the Loebner prize (www.loebner.net/Prizef/) of $100,000 and a gold medal remains unclaimed. However, a more achievable target is to see whether a computer can achieve the same level of performance as a human expert in a complex but bounded field. This became the aim of a young research student, Edward Shortliffe, in the early 1970s, looking for a suitable topic for his PhD thesis. A paper by Roberts and Visconti (Am J Hosp Pharm 1972), which highlighted the relatively poor performance of hospital doctors when diagnosing bacterial infection and recommending suitable treatment, suggested to him the goal of developing an ‘intelligent’ computer system which could do better.
The result was MYCIN - not the earliest successful expert system (that honour belongs to DENDRAL)but the first to demonstrate the use of production rules, a generic inference engine, certainty factors and comprehensive explanation and justification. MYCIN became not just a project but a programme lasting for over ten years and laying the foundations for much of the present day knowledge-based expert system technology.