[ExI] Watson On Jeopardy

spike spike66 at att.net
Tue Feb 15 17:27:36 UTC 2011


>... On Behalf Of Richard Loosemore
...

>...Apart from various bits of peripheral processing to catch easy cases, to
look for little tricks, and to eliminate useless non-content words, etc
etc., that is all it does.

>...It is a brick-stupid cluster analysis program....Richard Loosemore


Sure but that in itself is enormously educational.  The chess world was
rather shocked to learn how the best chess programs were disappointingly
simple.  All manner of tricky positional evaluation algorithms were tried,
but in the long run, they were overwhelmed by brute speed and simple
evaluation algorithms.  Today the best chess algorithms are not very
complicated.  What this taught us is that the best chess players are far
more stupid than we realized.  Chess is a simple game.  It only looks
complicated to simple-minded creatures such as humans.

In that sense I am not particularly surprised to learn that Watson is a
fairly simple program, but delighted in a sense.  That is the outcome I
wanted.  We have learned the magic of simple algorithms to do interesting
things.  The reason this is desirable is that simple algorithms are
accessible to more humans, which means we will write more of them to do such
things as watch us in the kitchen and tell us how to do the next step in
creating a meal for instance, or watch us working on a motorcycle and coach
us along.  Or teach our children.  Or do simple medical diagnoses by noting
each day our weigh (sensors in our computer chair) and sniffing the air
about our corpses and doing chemical analysis while we do our normal
activities at the computer.  They could watch what we eat and in what
quantities, and make annoying suggestions for instance.

Simple algorithms can do much for us.  Furthermore and more importantly,
simple algorithms can run on simpler processors.  This will likely be
enormously important as we progress to have vastly more numerous even if
simpler processors.

spike










More information about the extropy-chat mailing list