June 25, 2012 12:25 PSP Book - 9in x 6in 04-Junichi-Takeno-c04
Embodied Cognitive Science and Symbol Grounding Problem 59
Embodied cognitive science is a belief that intelligence can only
be understood in combination with physical systems. This concept
captured the hearts of many researchers because it specifically
proclaims that intelligence is in need of an embodied mechanical
system. Embodied cognitive science represented a large flow of
research covering the major achievements of robot researchers in
the 1980s, which was later summarized by Rolf Pfeifer, Federal
Institute of Technology Zurich, Switzerland, and other researchers
(Pfeifer, 2001).
Pfeifer and other researchers considered that the study of
artificial intelligence and robots had been critically blocked by two
major obstacles, and he tried to achieve a breakthrough using a new
keyword: embodiment. These obstacles are the frame problem and
symbol grounding problem.
“Frame” refers to a symbolically represented knowledge data-
base, and the term was originally proposed by Marvin Minsky. He de-
clared that human intelligence could be realized on a computer using
symbolic inference within two or three years. Certainly, artificial
intelligence using symbolic inference contributed to the develop-
ment of the MYCIN medical expert system designed to examine and
diagnose infectious diseases and a robot simulator called SHRDLU
(previously mentioned), which performed intellectual tasks as
instructed by humans although in the world of building blocks.
In 1997, the IBM research group developed Deep Blue, a chess-
playing computer. Readers may remember that the machine played
chess with world champion Garry Kasparov and won a five-game
match by two wins, one loss, and two draws. Deep Blue symbolized
and processed the chessboard and the positions of the chess pieces
and anticipated the next moves of its opponent by symbol-based
reasoning.
There is no doubt that symbolic inference is one of the
basic techniques for achieving artificial intelligence, but the frame
problem prevented the results for use in the future possibilities of
artificial intelligence.
John McCarthy is said to be the first person to point out
the frame problem (1969). The essence of the frame problem
is this: It is an illusion to believe that the knowledge database
describes everything, as it is obvious that knowledge databases
cannot describe the entire mass of knowledge. A nuclear reactor