Robotics Reference
In-Depth Information
dom might be helpful. Berliner had observed in Chess that the best pro-
grams were not those which emulated the selective thought processes of
very strong human players but those that used “brute force” methods—
examining many millions of chess positions in the analysis of a single
move. The fact that such programs can make moves as strong as those
from the world's top grandmasters is what Berliner means by saying
“...at best anthropomorphic solutions”—the moves were those that a
strong human player would make and therefore they could pass for be-
ing humanlike. The fact that the way to ultimate success at the chess-
boardlayinmethodsthat“...failedtocapturetherealgistofahuman's
method” was a sadness for Berliner because, as a very strong Chess player
himself, he had long tried to succeed at Chess programming by emu-
lating the highly selective analytical process of the human Chess mind,
whereby in most positions only a small fraction of the available moves
are analyzed further. Instead he was reluctantly compelled to admit that
a “brute force” approach in Chess programming had proved superior to
the more intelligent “selective” methods of Chess analysis that he and
a few other researchers in this field had promoted. And as it has been
in Chess, so it has also been in other areas of AI, where success has of-
ten come, not by emulating human methods but by devising alternative
approaches that rely very much on the availability of extremely fast com-
puters, often with very large memory capacities.
The perspicacity of Berliner's statement underlies the conviction of
those futurists who believe very strongly that the exponential growth in
computer power will be a cornerstone of the growth in Artificial Intelli-
gence research and its achievements during the first half of the twenty-
first century.
What Is Exponential Growth?
Different things grow at different rates. Some things grow at a constant
rate (linear growth), for example a tree that grows a certain number of
inches in a year. Some things grow at what is called a geometric rate,
which means that the growth during a particular period is found by mul-
tiplying the size at the start of that period by a fixed number, for example
doubling the number of transistors per year that can fit on a fixed area of
silicon. An exponential rate of growth is one for which, at any point in
time, the rate of growth depends on how much (of whatever) is already
there at the time—the more there is, the faster the rate of growth.
Search WWH ::




Custom Search