DENDRAL Project Begins at Stanford

Much of this Timeline's history of AI will focus on the cybernetic-connectionist-deep learning thread. However, we will from time to time highlight major advancements in symbolic, expert systems. One of the earliest hard-coded AI expert systems used for scientific reasoning was DENDRAL.

In 1965 at Stanford University, computer scientist Edward Feigenbaum asked Nobel laureate Joshua Lederberg for a scientific problem to study hypothesis formation. Lederberg, involved in NASA’s search for life on Mars, suggested identifying unknown molecules from mass spectrometry data.

And they did and they created DENDRAL, a program that generated every chemically valid arrangement of a molecule’s atoms and pruned the list with chemists’ "expert" rules of thumb. The team asked chemist Carl Djerassi, “What do you know about this problem that the program does not know?” Bruce Buchanan and Georgia Sutherland turned the answers into rules, working with Djerassi’s colleague Alan Duffield.

It was precisely this kind of expert system that would gain in popularity in the '70s and '80s while neural network AI experienced a prolonged winter.