文档介绍：Chapter40
The Appeal of ParallelDistributed Processing
JamesL . McClelland, David E. Rumelhart,
andGeoffrey E . Hinton
What makes people smarter than machines? They certainly are not quicker or more
precise. Yet people are far better at perceiving objects in natural scenesand noting their
relations, at understandinglanguage and retrieving contextually appropriate information
from memory, at making plans and carrying out contextually appropriate actions,
and at a wide range of other natural cognitive tasks. People are also far better at
learning to do thesethings more accuratelyand fluently through processingexperience .
What is the basisfor thesedifferences ? One answer, the classicone we
" " perhaps might
expect from artificial intelligence, is software. If we only had the puter
program, the argument goes, we might be able to capture the fluidity and adaptability
of human information processing.
Certainly this answeris partially correct. There have beengreat breakthroughsin our
understanding of cognition as a result of the development of expressive high-level
computer languagesand powerful algorithms. No doubt there will be more suchbreakthroughs
in the future. However, we do not think that software is the whole .
' story
In our view, people are smarterthan today putersbecause the brain employs a
putational architecturethat is more suited to deal with a central aspectof the
natural information processingtasks that people are so good at. We will show through
examplesthat these tasks generally require the simultaneousconsideration of many
piecesof information or constraints. Eachconstraint may be imperfectly speci Bed and
ambiguous, yet eachcan playa potentially decisiverole in determining the e of
processing. After examining thesepoints , we will introduce putationalframework
for modeling cognitive prpcessesthat seemswell suited to exploiting theseconstraints
and that seemscloser than other frameworksto the style putation