🚀 go-pugleaf

In article <2p14e0$3j1@louie.udel.edu>,
Wendy McKnitt <mcknitt@hercules.cis.udel.edu> wrote:

> Let me know if your AI group there needs anyone with some
>expertise in Natural Language Processing/Generation...I'm gonna be needing
>a job come next December! ;)

Excellent! We are *very* interested in having your join our research team!
As you know, we are working on creating AI posting programs to simulate
real human beings. 

However, as you can tell, we do have a problem parsing "natural" language. 
That is why so many of the posts coming out of DTCC seem to have that
"gawd, don't they have a single clue" quality about them.  We try and
cover up the rough (ie, mis-parsed) items by simply rearranging a few
words, and spitting out the stuff so as to make it *sound* like it was
understood. 

A good example of a very early (and crude) case of this was feeding the
output of "Zippy the PinHead" into MIT's Eliza program. 

What we hope to accomplish someday is to "simulate" a Usenet feed
by running a bunch of daemons that can generate up to 100 megabytes of 
total utter bullshit across a few thousand groups originating from a few 
hundred thousand pseudo-individuals.    We can then sell our program to 
institutions all over the world who are eager to simply cut their Usenet 
feeds, but fear a backlash from their user community.

With our system, any institution or company can cut their Usenet 
connection and replace it with our pseudo posting software and their user 
community will be none-the-wiser!

Ideally, this pseudo posting software can work in real time as each local
user reads news. For example, a user runs a reader like trn, a random (but
believable) number of unread article count is presented to them for a
given group. As they enter the group, the AI poster software generates a
number of subject titles, some remembered from previous session, some new
based on new topics created by parsing current events (perhaps from a
stolen Clarinet feed off the net, for example.) Then as a user enters a
thread, a "tree" of articles is built in real-time. As the user follows
this fake thread, the software will be working on composing the next
followups the user is likely to read. 

With this feature in place, a site can simulate a 5 gig news spool 
without having to use any disk space to store it!!!!!

(Hmmm, come to think of it, a stolen Clarinet feed really isn't needed, 
since the vast majority of usenet readers don't have a clue about the 
outside world anyway. We can just "make up" current events to create 
posts about, sort of like how the Christian Coalition does when they send 
out their news flashes...)

There *are* some difficulties, however.  The usenet posting volume is
growing at a rate of about n*logn lately. Clearly a problem in the making.
The users may expect that much more useless horse shit to be flowing into
their systems.  Believe it or not, spewing out this much mindless garbage
*does* take a lot of CPU power, so some ideas we have come up with center
around throwing the user community off so they don't notice. For example,
have the posting programs start rumours that the "death of usenet"
predictions are coming true, or else just create more robo-mass-posters
like the "Green Card Lottery" posts to fill up spool space. These should
create enough flame wars amongst the local user community to help fill up
any slack. Another possibility is to simulate floods of clueless newbies
coming onto the net, like fake CI$ and thousands of BBS'es finally getting
their act together and dumping millions of clueless dweebs onto the net
posting stuff like "hey, this is cool. Is anyone getting my post?"

So, Wendy, as you can tell, this is a very rewarding, and much needed 
project for the "good of all humanity."   Hope you can join us someday!
-- 
Ken Weaverling          weave@dtcc.edu |*| My opinions .NEQ. college's position