Tony Harkin noted that Doron Zeilberger automated the
discovery of combinatorial identities and has long been an advocate of
using computers to find new theorems. Zeilberger calls his computer
"Shalosh B. Ekhad" and has given it co-authorship on papers!
See this Wikipedia page and this essay by Zeilberger and Ekhad's home page at Rutger's University.
Siemion Fajtlowicz developed the program Graffiti in the 1980's. Here is a
summary of our exchange:
The claimed computer rediscovery of Newton's Laws and the implied promise of authentic automated discovery in the
physical sciences is not new. The first such claims were made in the seventies by Herbert Simon and his colleagues
who wrote a program named "Bacon" (after Francis Bacon).
Similar claims about mathematical discoveries were made by Douglas Lenat about a computer program
Artificial Mathematician. This program supposedly rediscovered the Golbach conjecture, one of the most famous in mathematics,
according to which every number larger than 4 is a sum of two primes. However, neither of these
programs were capable of making authentic discoveries.
In mid eighties, after learning about "Bacon" (I've learned about "AM" later) I began writing the program "Graffiti".
This program generated conjectures that within a few years led to dozens of publications. Although Herb Simon,
acknowledged that "Graffiti" became the first computer program that authentically
discovered hypotheses of interest to mathematicians, he adamantly opposed the idea that the same could be done in the physical
sciences. Yet a few years later Graffiti did make such conjectures in chemistry. In 2001, I announced a hypothesis, for all
practical purposes due exclusively to "Graffiti" about classical stable fullerenes. Although this conjecture was initially
questioned, it was subsequently proved and has inspired further work since.