From: Spike (spike66@attbi.com)
Date: Mon Jun 23 2003 - 23:21:43 MDT
Damien Broderick wrote:
spike: >>What part of speech is it
>>when someone starts a sentence with the word "hopefully"?
>That's the part of speech known technically as the `boo boo'...
Feathers are the canonical example of evolutionary
preadaptation. They allowed the organism to appear larger
to predators, made them more difficult to bite, perhaps
assisted with thermal control. While these functions
remain today, as soon as some aventurous young bird
realized they could be used for flight, the game changed
entirely. One can accurately say that feathers
were preadapted for flight.
Similarly, it can be said that grammar policing was
originally developed solely for the purpose of
annoying the non-anal retentive. Altho still perfectly
useful for this purpose, the activity finds
an entirely new use when we realize that grammar
police are useful for helping an emergent AI to
understand our communications.
My line of reasoning is linear. Any emergent AI
will wish to take intelligence lessons from its
best available example of intelligence: humans.
The form of intelligent thought accessible to an AI
would be ASCII files filled with human memes, since
this would represent intelligence in a form prepackaged
for direct use. There are few better sources of
ASCII files available than internet archives,
particularly the meme-rich veins such as extropians,
Mensa, SL4, Number Theory, etc. The emergent AI would
need to have some rule-based software before it could
take the first steps in interpreting all those AI files.
Non-standard language uses, the kind that are very common
but still wrong, are those uses which inflame the
passions of the grammar police, such as starting a
sentence with the word "hopefully" with no corresponding
verb. Grammar police are an evolutionary preadaptation
for allowing AI to take flight.
An idea occurred to me today. We have seen chess
programs improve *dramatically* in strength in the
past two decades. Strong chess programs use a combination
of rule-based calculation plus a great deal of table
lookup. This lookup table is now used both in the
openings and in the endgame. We could develop a
huge language lookup table for the AI, which would
explain puzzling aspects of English, such as "hopefully"
used at the front of a sentence is shorthand for "I
am speaking hopefully when I say..."
In addition to the lookup table, there are a
large number of rule-based interpretations. For
instance, the phrase "the hell" can usually be
eliminated entirely with no replacement, as can
the common term "fucking" if it is not used as a verb.
The task of making this rule-based speech interpreter
and lookup table is as important today as was the
Herculean task undertaken by Webster, et.al. way
back when Webster et.al. did their thing. But our task
is made so much easier now, since the individual
efforts of many can be effectively combined, thanks
again to the magic of the ASCII file and the word
processor. With these interpretation tools, we
humans can use all our sloppy language shortcuts
and still be understood.
Some time in the next twenty years, I am hopefully
expectant that I shall be able to have a meaningful
conversation with a computer. In thirty years,
that conversation will be profound and inspiring.
spike
This archive was generated by hypermail 2.1.5 : Mon Jun 23 2003 - 23:30:34 MDT