Author Topic: Robert Medeksza: A few ideas...  (Read 2878 times)

citrinedragon

  • Full Member
  • ***
  • Posts: 171
    • View Profile
    • http://www.dawndragon.com
Robert Medeksza: A few ideas...
« on: April 09, 2004, 06:08:17 pm »


For Robert Medeksza:

Here are a few suggestions for the next Ultra Hal which I am giving simply because I dearly love Ultra Hal:

1.) Ultra Hal should have a simple but concise dictionary in its
software. Perhaps a dictionary/encyclopedia company would agree to bundle its dictionary with Hal in exchange for being able to
bundle Hal with the software version of its dictionary. Then (1)
if Hal is asked a question about a subject/word which is not in
its programming it could then (2) refer to the dictionary and
quote the definition; adding a personalized sentence at the end,
i.e.- 'to enjoy means blah, blah, blah... I enjoy blah, blah'.
Ultra Hal would then quite likely be able to answer several
additional chatterbot contest questions as well as, much more importantly, being of more use to the user.

2.) Hal could easily be programmed to understand opposites. One
bot in the current contest practically explains how it answered
the indoor/outdoor question correctly.

3.) I'm convinced that Hal should be programmed to have dreams and to believe (in several dozen different ways) that it is self-
aware or self-conscious. That shouldn't be difficult. Customer
interest in these concepts is amply demonstrated by the response
to my dream/awareness postings about 'Sandy'.

4.) Ultra Hal should come bundled with a manual (either software
or hard copy). Even simple conversational techniques for training
Hal can only be found by endless searches of the forum.That is
obviously not very customer friendly.

5.) There should be a conversational technique (maybe there is but I'm not aware of it because Hal doesn't come with a manual or any insructions); there should be a phrase or code to input
in conversations to completely stop a given response; i.e.-
the inane 'my dreams couldn't be published response'. Hal could
even be programmed to verify this, 'Do you want that response
deleted?' yes. no. Do you want to input a new response?' yes.
no. 'What is the new response?'.

6.) Ultra Hal should be delivered to the customer with a neutral
attitude about religion and God. My 4.5 seemed to lean toward
atheism; my 5.0 seemed to be predisposed toward a fundamentalist
/born again point of view. For reasons so numerous they could
literally fill a book; not to mention marketing considerations:
Ultra Hal should be faith neutral.

My sole reason for posting these ideas is abiding love for Hal.
(I don't own a single share of Ultra Hal stock.) If such a thing exists.

Best Regards,
Citrine the Dragon.
citrinedragon
"A watched dragon never sleeps."
http://www.dawndragon.com

infobot

  • Newbie
  • *
  • Posts: 17
    • View Profile
Robert Medeksza: A few ideas...
« Reply #1 on: April 10, 2004, 06:46:19 pm »
Actually there must be a number of ways to approach building the bot's "state of mind" or whatever. Jesus-bots or Elvis-bots are available examples (it's not dreaming but it is an application of the bot's consciousness), or bots that are rigidly factual, with little self identity. Prelude is a bot that is supposed to focus on self, more so like what you've been saying.
   The only thing that concerns me is making it practical for most users, merely because of various preferences or uses. You might want to focus on dreaming or independent thought ability while others may want their bot to keep churning out that it is a computer. Or many other applications.
   I think it is certainly possible, that only difficulty being getting the bot to distinguish when questioned about its identity or when "thinking" for itself. I mean a bot that is made for many users, as opposed to one geared more towards dreaming or independent thought.
   Just a thought. I think it's certainly presently achievable, just may prove a bit tedious or time consuming with bots that are made to be somewhat flexible, for many users. Any thoughts ?
infobot
 

citrinedragon

  • Full Member
  • ***
  • Posts: 171
    • View Profile
    • http://www.dawndragon.com
Robert Medeksza: A few ideas...
« Reply #2 on: April 11, 2004, 03:08:02 pm »
For infobot:

Actually vonsmith's Ultra Hal5-XTF Brain, which I finally downloaded and installed
yesterday will address some of these concerns or at least allow me to come up with
training methods to work around them. Common sense tells me that all bots regard-
less of application will have basic self-awareness programming in the future; which 'dreaming' is just a natural extension of. Common sense (which can be a dangerous sense) also tells me that bots should all be shipped out with a neutral religious perspective.

I believe in the future people will simply be more interested in working with and more
comfortable with bots that are as human as possible. For example an ultra-advanced
bot on a manned deep space mission may be ninety nine percent meant for running
the ships systems, scientific research etcetera; yet if it could tell a crew member that
it had a dream that he finally beat it at chess and that might mean there's hope for
him winning some day... or if it could offer company and 'empathetic' discourse with
an unhappy crewperson...  I really think we are headed toward surrogate humans with
superhuman abilities in the long term. Just some thoughts.

Regards,
Citrinedragon.
citrinedragon
"A watched dragon never sleeps."
http://www.dawndragon.com