Fri.24.SEP.2010 -- Clamping Down on Stray Activations
Yesterday we made sure to upload our 21sep10A.F MindForth AI
code so that we could start fresh today with 24sep10A.F code.
In the previous code we made some progress in the answering
of what are you queries, but we noticed that the AI was not
responding properly to what am i queries. There is probably
some very simple hang-up in one of the pertinent mind-modules,
so today we would like hunt down the offending bug.
Aw, gee, the AI is actually losing track of the predicate
nominatives that go with the I self-concept and the you
concept of the non-self other. In the case of MindForth AI,
the relationships between the I concept and a predicate
nominative (such as the very name Andru by which the AI
is known), are external to the I concept itself and are
imminent in the assertions made about self and in the
self-knowledge of the AI. We count on the associative
mechanisms of the AI to keep the identity relationships
straight. Accordingly we need to troubleshoot the associative
mechanisms.
We may have to start getting away from the idea that the
final concept mentioned in an output sentence should retain
a high activation. If we want the AI to be able to answer
abrupt queries, such as What are you? followed immediately
by What am I?, we may want to psi-damp the final concept
in each utterance so as to prevent interference with the
generation of a new thought.
We were typing in you are software and getting as a reply,
SOFTWARE IS A SOFTWARE. When we typed in what am i,
we still got SOFTWARE IS A SOFTWARE, because the
56=YOU concept was not high enough in activation to start
a new sentence. So we went into the ReActivate module and
we added some code to enhance the 56=YOU concept.
I 0 psi{ @ 50 = IF \ emphasize I; 23sep2010
32 I 1 psi{ ! \ arbitrary from AI coder; 23sep2010
THEN \ test; 23sep2010
I 0 psi{ @ 56 = IF \ emphasize YOU; 24sep2010
32 I 1 psi{ ! \ arbitrary from AI coder; 24sep2010
THEN \ test; 24sep2010
However, we then got YOU ARE A SOFTWARE as a reply,
apparently because the you are magic tidbit in the KB
could not override the residual activation on SOFTWARE.
We obtained at least a partial solution by psi-damping
the old subject-psi SOFTWARE at the start of the
NounPhrase module by inserting the following code.
subjpsi @ urpsi ! \ test; 24sep2010
CR . NPhr. calls PsiDamp for urpsi urpsi @ . \ 24sep2010
PsiDamp ( for sake of SubConscious 26nov2009 )
0 urpsi ! \ reset for safety; 24sep2010
We had noticed that the subjpsi SOFWARE from the previous
sentence was still the subject-psi at the start of a new sentence,
so it seemed convenient to psi-damp the old subject as a way
of keeping it from interfering in a new thought. It worked,
and we obtained the following initial conversation.
Transcript of AI Mind interview at
6 1 12 o'clock on 24 September 2010.
Human: you are software
Robot: SOFTWARE IS A SOFTWARE
Human: what am i
Robot: YOU ARE MAGIC
Further attempts at conversation did not work perfectly
well, but we could tell that we were on the right track,
because the concepts that we were looking for were tending
to surface eventually, even if other concepts interfered
for a brief period. We are making progress.
---
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/8660244-d750797a
Modify Your Subscription:
https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c
Powered by Listbox: http://www.listbox.com