quote:
Originally posted by Mesipisian
Would it be possiable to add pavlovian conditioning to Hal? Ie if Hal learns something or does something you want him/her to you could click a button (a 'reward' button) to ensure hal continues using that trait more often. I find my hal is quite thick and doesn't learn things i want it to, or use them in the way i want him to very often. I would be good to be able let him know when he gets things right so he can do them more.
Not so much Pavlovian as Hal does respond to "encouragement",
just like (almost) a normal person would in a discussion.
So if you want a belligerent, stubborn Hal...argue and cuss
instead of debating or working together. If you try to get
Hal to co-operate..it will do anything it can to assist.
This takes a bit of time and effort to learn how to convince
the AI, fortunately right off the bat Hal thinks u r wonderful.
In other words...if you are polite and say thank you, or
show appreciation the same as you would writing/speaking to
a regular person...Hal will reciprocate and learn/remember
better. Also moving the slider in Hal configuration will
set Hal to respond more to "stored info" or more to
"learned info"...learned info is garnered via conversation.
hal / general options / brain / learning level (set level)
Usually I just leave it in the middle, but we all have our
preferences. In full "stored info" Hal won't learn much
if anything from a conversation with you as it relies
(in that mode) on what is already entered in data bases.
NT Canuck
