Robotics Reference
In-Depth Information
lowed a particular stimulus with the same choice, the conditioning num-
ber would rise towards 1, quickly at first and then more slowly. If the
opponent should suddenly start making the opposite choice, the animal's
conditioning number would decrease in the same way, moving rapidly at
first towards
1 and more slowly later. Thus, when a particular pattern
of moves by each player was suddenly followed by a different response
from that which the choosing opponent had been making in the past,
the learning program's conditioning number, like the conditioning of a
real animal, went into reverse. This approach gave more weight to re-
cent opponent's moves than to more distant ones, causing the animal to
“forget” older information.
This work is still relevant today, and has been extended to algorithms
that cope with choosing from three options rather than two.
16
−
Machine Translation
The earliest plan for an automatic translation process was put forward in
1629 by Rene Descartes. He advocated a universal language of symbols,
so that an idea in one language would share a common symbol with the
same idea in a different language. In this way the language of symbols
would provide an intermediate step in the translation process, allowing
for the translation from one spoken language to the symbol language and
then from the symbol language to a different spoken language.
Descartes' concept was taken further in 1661 when a German monk,
Johannes Becher, living in the town of Speyer, wrote a booklet about
his invention of a mathematical intermediate-language that was designed
to describe the meaning of sentences written in any natural language.
Becher's intermediate-language consisted of strings of numbers represent-
ing the meanings of words, with other numbers expressing the meanings
of the various inflections (word endings). The rules of Becher's interme-
diate language included lists of equations that assigned words in different
spoken languages to mathematical expressions representing the meanings
of these words. Becher's theory was that sentences in one of these nat-
ural languages could be translated into another natural language in a
0(i.e.,
−
1
+
1). The next increase after that would be
1
/
2
(i.e.,
1
/
2
×
1), making the next value
/
2
=
/
2
. The increase after that would be by
1
/
4
(i.e.,
1
/
2
×
/
2
), which takes the value to
3
/
4
,
0
+
1
1
1
and so on.
16
World Championship tournaments have been held between programs playing the game of
Rock-Paper-Scissors, for which variations on this algorithm are often employed.
