The “Modern” History of Artificial
the Dartmouth Conference
artificial intelligence is "making a machine behave in ways that would
called intelligent if a human were so behaving". Looking
back on this definition, people today
seem to disagree with it since it ignores the possibility of achieving
this definition of artificial intelligence because it does not rule out
possiblilty for strong AI. In fact, it
does leave the possibilities quite open.
Artificial Inteliigence is "intelligence
arising from an artificial
many many different definitions fo AI. Most
definitions could be categorized as concerning either systems that think like humans, systems that act like
humans, systems that think rationally or systems
that act rationally. (Wikipedia.com)
So as you can see, it is
with a truly precise definition as to what exactly artificial
in modern terms.
is a look at some early AI programs that helped pave the path into the
day for artificial intelligence and robotics:
first working AI
written in the UK
by Christopher Strachey, Dietrich Prinz, and Anthony Oettinger.
he was also an amateur
programmer. Later Strachey became
Director of the Programming Research Group at Oxford University.
Prinz worked for the engineering firm of Ferranti Ltd.
This engineering firm would become famous for
building the Ferranti Mark
I computer in collaboration
with Manchester University
(This computer is the
computer that contained the earliest artificial intelligence programs,
ran them). Oettinger worked at the Mathematical Laboratory at Cambridge University,
home of the EDSAC computer.
(Ferranti Mark I computer)
Strachey decided that the
checkers (a.k.a. draughts) would be ideal for creating his first
could play a game. In May 1951, Strachey initially coded his checkers
for the pilot model of Turing's Automatic
Computing Engine. His
efforts were unsuccessful as the machine
did not work. Coding errors and hardware
changes were what defeated led to the demise of his program. Strachey was very dissatisfied with the
method employed in the program which evaluated board positions. He
forward, using his dissatisfaction to fuel his creativity, and wrote an
version for the Ferranti Mark I at Manchester. By
the summer of 1952, his new
could "play a complete game of Draughts at a reasonable speed", said
Prinz's chess program, also
the Ferranti Mark I, first ran in November 1951. The program would
every possible move until a solution was found. On average several
moves had to be examined in the course of solving a problem. For this reason and the fact that the
processing speed for computers was pretty slow back then, the program
take a very long time to choose the move in which it should make.
Turing started to program
what he called
his “Turochamp chess-player” on the Ferranti Mark I.
What made this program different was that the
Turochamp could play a complete game and operated not by exhaustive
under the guidance of rule-of-thumb principles devised by Turing. Unfortunately, he was not able to finish this
first AI program to run in the U.S.
was also a checkers program, written in 1952 by Arthur Samuel of IBM
IBM 701. Strachey had publicized his program at a computer conference
Using this, Samuel worked over a period of years to come up with a
faster program. In 1955 he added features that enabled the program to
from experience, and therefore improve its play. This
was key to the eventual progress that
would be made in computing programs and artificial intelligence. Samuel
included mechanisms for both rote learning and generalization. Samuel made continuous improvements to the
program until it finally reached the point in which it beat a former Connecticut
champion in 1962.
- About 19 IBM 701
built from 1952 to 1955. Most of these were sent to government agencies
defense, atomic research, navy and the weather bureau.
To further enhance his
added a type of
learning called heuristics. This form of
learning utilized the fact that two programs could play against each
if one program beat the other, the program who was defeated changed
tactics to match the one that beat it. This
seemed to be biological evolution (except with no
biology). In essence, the smarter survived
and only that
program would carry on.
(The History of Computing. http://www.alanturing.net)
language communication: Is it possible?
programs called Eliza and Parry were invented. Parry, written by Stanford University psychiatrist Kenneth
simulated a human paranoiac. Parry's responses are capitalized in the
extract from a "conversation" between Parry and a psychiatric
Psychologist: Why are you in
Parry: I SHOULDN'T BE HERE.
Psychologist: Who brought
Parry: THE POLICE.
Psychologist: What trouble
did you have with the
Parry: COPS DON'T DO THEIR JOB.
Psychologist: What do they
Parry: COPS ARREST THE WRONG PEOPLE.
Psychologist: They arrest
the right ones
Parry: THE LAW SHOULD GET THOSE ITALIAN CROOKS.
Psychologist: Are you
Parry: MY BACKGROUND IS BRITISH BUT I WAS BORN IN THIS COUNTRY.
Psychologist: Do you know
any Italian crooks?
Parry: I TRY TO AVOID THE UNDERWORLD.
What is amazing about these
few lines taken from
an interview was that psychiatrists were asked to say whether this was
paranoic person or a computer,
and some of the psychiatrists thought it was an actual person.
Eliza was written by Joseph
Weizenbaum at MIT in
the same year as Parry. Eliza was
supposed to represent a human therapist. Read
the lines and you will be amazed at the life-like
Man: Men are all alike.
Eliza: IN WHAT WAY
Man: They're always bugging
us about something
Eliza: CAN YOU THINK OF A SPECIFIC EXAMPLE
Man: Well, my boyfriend made
me come here.
Eliza: YOUR BOYFRIEND MADE YOU COME HERE
Man: He says I'm depressed
much of the time.
Eliza: I AM SORRY TO HEAR THAT YOU ARE DEPRESSED
Man: It's true. I am
Eliza: DO YOU THINK COMING HERE WILL HELP YOU NOT TO BE UNHAPPY
Man: I need some help, that
much seems certain.
Eliza: WHAT WOULD IT MEAN TO YOU IF YOU GOT SOME HELP
(The History of Computing. http://www.alanturing.net)
Both Eliza and Parry were
deemed to have
the intelligence of a "juke-box", as one psychologist explained. They both used simple techniques of word
recognition and play-back. But even so,
I think these two programs are pretty neat for their time.
Learning in AI:
Learning is distinguished
into a number of
different forms. The simplest is learning by trial-and-error. The
simple memorizing of
solutions to problems, words of vocabulary, etc…is known as rote
learning. Rote learning is
easy to perform on a computer. However,
generalization is hard to work into a computer program.
What I mean by generalization is that the
program will be able to come up with a solution when a problem is faced
has not been previously tackled. Ex: If
the computer program was used to add simple numbers, if it had not been
introduced to 2+4=6 previously, it would have a hard time coming up
correct answer, 6.
GA – Genetic
Algorithm : More than
modern, perhaps future
GA is the first in a field
that is called
"evolutionary computing". It
was introduced by a man named John Holland in 1975 in collaboration
research group based out of the U.
of Michigan, Ann Arbor. GA's are
similar to [the concept Darwin
gave us] natural evolution. This type of
computing produces successive generations of software that increasingly
better and better for their specific goal(s).
Current use of a GA system
can be seen in
detective work. A witness in
coorperation with a GA system produce a face that becomes increasingly
to the face of the criminal that the witness recollected.
(The History of
I have found a
flowchart that will help you understand the entire process that takes
a GA system
taken from http://www.sv.vt.edu/classes/ESM4714/Student_Proj/class94/