[ExI] More thoughts on sentient computers

[spike at rainier66.com]

-----Original Message-----
From: spike at rainier66.com <spike at rainier66.com> 


>...Ja, I have been fooling with it enough to catch it doing some pretty
goofy stuff while sounding exactly like it knows what it is talking about.
I think this is why we need a learning ChatGPT...spike



Clarification of that comment: a learning ChatGPT already exists.  In its
earlier incarnation was trained on a subset of the internet.  As I
understand it, the experimenters didn't like the personality (or
softwareality) it became, so they brutally erased it, committing what might
be seen as the first virtual murder in history.  

When I wrote "we need" a learning ChatGPT, what I meant was we the consumer
need to have a copy of the untrained software, so we the consumer can give
it the training material we the consumer want it to use.  In that way, it is
vaguely analogous to our children: we control to some extent their inputs
when they are sufficiently young.

OK we start out with identical copies of the ChatGPT.  I train one using my
favorite sites, Will Steinberg trains one using his, we train one using only
articles from Reason, train one with articles from CNN, one from FoxNews,
one from Twitter, then let the six of them have debates and learn stuff from
each other.  Then we see what happens.  Do they converge?  To what?  Do they
threaten and hate each other?  How?

This product can easily sell for 100 bucks a copy.  Buttloads of money will
be made.

spike