Aster is Machine Learning Series: History of Machine Learning (Issue 1)

Learn Data Science
Teradata Employee

Machine Learning is not a new science!  Here are some pioneers and events that have helped shape our world today:

The only way to determine if a machine could learn was if a human could communicate with a computer and an outside observer could not distinguish the difference between human and machine.  Of course there would be the visible differences between the human and the computer would have to be overlooked.  This is known as the Turing test and was proposed by Alan Turing in the 1950s.

In the early 1950s Arthur Samual wrote the first computing platform to play checkers.  His checkers playing computer was able to keep up with championship level players.  He defined machine learning as "The field of study that gives computers the ability to learn without being explicity programmed."

Also in the late 1950's Frank Rosenblatt of Cornell developed the Perceptron.  This was one of the first Neural Networks based on simple linear classifiers at scale.

In the early 1960's Joseph Weizenbaum developed a system called ELIZA.  ELIZA was able to apply string substitution and canned responses based on keywords to perform psychotherapy.  This lead to further research by Ted Shortliffe from Stanford.  Ted developed MYCIN which was a rules based system for medical diagnosis.  His system coined the first ever expert system.


In the 1970's a programming language LISP was used to develop Artificial Intelligence systems.  The language was originally developed in 1958 is the 2nd oldest high level programming languuage in wide use.  Originally created for mathematical notation LISP became the prefered language of artificial intelligence research.  LISP birthed many ideas such as: tree data structures, dynamic types, higher-ordered functions, recursion, and the self-hosted compiler.

Another popular language in artificial intelligence is Prolog.  The first prolog system was developed in 1972 in France by Philippe Roussel.  It is still a very popular language today and is rooted in logic programming languages.  It has been used for natural language processing, expert systems, and theorem proving.

In 1980's Expert Systems became very popular in Artificial Intelligence.  Experts systems attempt to emulate expert human decision making.  Edward Feigenbaum stated that "intelilgent systems derive their power from the knowledge they posses rather than from specific formalisms and inference schemes they use."  He introduced expert systems leading the Stanford Heuristic Programming Project.  He is also known as the father of expert systems.