Messages

16

Programming using the Ultra Hal Brain Editor / Re: The concept of Now

« on: October 17, 2012, 11:55:46 am »

I'll post here, rather than start a new topic, although HAL still has problems with time.
  'Now' is used as an adverb, like 'today', but these are really time pronouns.
'now' means 'at this moment'
'today' means 'at this day'
HAL takes all adverbials of time as triggers for the scheduler.
It needs to resolve the pronouns before doing so, otherwise the pronoun keeps getting
re-evaluated, so that tomorrow is still today, now is still that old now, etc.

17

General Discussion / Re: A.I Logic theories.

« on: October 09, 2012, 05:18:39 pm »

Ouch - I tried a reply last week but went over my login time limit and lost it..

Logical thinking requires a great deal of processing that comes so easy to us.

My A.I. handles statements in 3 steps:
 1. translation to mentalese  (this removes pronouns and determiners, leaving activated only the 'urls'
     of the valid subjects)
 2. verification of whether the info is already known
 3. integration: if not known, then add to memory (based on trust)

The memory is a list of lists, rather than a narrative.  For example:

subject: All thinking
 relation: may be logical
 relation: may require processing
  determiner: great deal of    (adjective unit_noun reference)
 function: if logical then require..

subject: processing of thinking    (possession of thinking)
 relation: come
 adverb: easy   (easily)
  intensifier: so
 dative: to us

If asked a question, then step 3 changes from integration to
 3. extraction: yes, no, don't know, or fill in query pronoun

18

Programming using the Ultra Hal Brain Editor / Re: learn from text

« on: September 26, 2012, 01:26:15 pm »

To the programmers,
  I'm surprised that HAL can't take tests.  That is a basic tool of learning for a brain.
It is a simple matter of re-direction to have questions come from a file, and answers
go to a file, instead of the HAL 'pad'.  In my own studies, I tell the AI to read a file
of questions, such as 'Read TEST.TXT' or 'Take TEST.TXT', after telling it to 'Record to TEST.REC'.
I also include the answers following the questions, so the AI can score itself, although
it doesn't do the self-scoring in mentalese yet, so nuances of articulation cause mismatches
that lower the score.
(PS, a '.' followed by a char is never a real 'period')

19

Ultra Hal 7.0 / Re: Is there a way to make Hal Talk on my Facebook chat for me?

« on: September 11, 2012, 12:17:56 pm »

Art, (see Carl2's late comment an animism..), I like the idea of seeding bots onto the web that garner resources to help their own evolution.
We know that HAL is web-able from the site demo, but how can users do it?

20

Ultra Hal 7.0 / Re: Is there a way to make Hal Talk on my Facebook chat for me?

« on: September 04, 2012, 05:04:53 pm »

jessessoulmate, putting HAL on facebook would devolve into bots friending bots, but is that so bad?
(maybe that's already what it is...)
Can someone post some java script - HALs should get in on some forums (woe to the spam checkers).

21

General Discussion / Re: A question on animism

« on: September 04, 2012, 04:37:54 pm »

'Soul' is a loaded word that I will avoid.
Mirror recognition is a problem related to photo/movie recognition, the difference being that mirrors give
positional feedback; ie, we recognize ourselves (even in a mask) by seeing that the figure copies our motions.
(refer to false mirror skits ala Harpo Marx)
Computer recognizers are only now becoming feasible. 
I think that means that the soul arises from complexity.  Avoidance voided!

22

Ultra Hal 7.0 / Re: NewB questions

« on: August 28, 2012, 04:33:21 pm »

Art, you kind of saw my point about time nouns, but not my conclusion:
  'at' is not needed, but then neither are any of these prepositions before time nouns:
ex: on Tuesday, in August, at 5 [o'clock], in a week...
So, does mentalese insert a preposition [such as 'at'] when it is missing;
  ex: I'll vote, Tuesday -> I'll vote, on Tuesday
or does it remove the redundant preposition when we use one
  ex: I'll vote on Tuesday -> I'll vote, Tuesday
I tend to believe the former, perhaps because different prepositions affect the
aspect of time nouns; ex: I'll vote in an hour  vs.  I'll vote for an hour.

Art, your comments about not ending questions with a preposition sound like
grammarian fussing - Brains are perfectly capable of understanding your 'bad'
examples, where a query pronoun is moved to the head leaving a trace following
the preposition.  The same goes for starting a sentence with an adverbial preposition.
PS: I hate the term 'prepositional phrase'.  I think of them as reference phrases,
and subdivide them into spacial and temporal references.

 What I wanted to know about HAL was: Can he identify time nouns in a copula
(a phrase with 'be', which requires equal animacy of subj and obj) and use that
to find the right antecedant of 'it' (3P vs 4P)?
PS: In my own studies, I use 'now' as the 4th person antecedant, converting it
from an adverb to a 4P noun (with an implied 'at').
  ex: what time is it -> what time is at now
I know that sounds clunky to us, but it logically resolves the pronoun.
     

23

Ultra Hal 7.0 / Re: NewB questions it

« on: August 14, 2012, 05:59:42 pm »

Thanks for the comments on 'it'.
  English has a 4th person, namely 'nature' or 'time', for which the 3rd person pronoun 'it'
is overloaded (used for more than one operation).
We detect the 3rd vs 4th person based on 'natural' or 'temporal' animacy.
ex: what time is it?  {'time' is temporal}
ex: it is raining.  {'rain' is natural}
ex: I will vote in the election.  It is Tuesday.  {this is ambiguous}
   Since Tuesday is temporal, 'it' could refer to the (4th P) current time,
   but 'it' could instead refer to the election, which is a (3rd P) thing.

The mental dictionary needs to include 'nature' and 'time' codes with each definition,
(in addition to 'abstract', 'physical', 'animate', 'human', ...).

I'll also mention here that most 'time' words are used as (adverbial) references, but we
often omit the preposition ('at') because we know that time is not a thing.
Thus we can say 'I will vote today' when we mean 'I will vote, today' or 'I will vote at today'.
If we used 'at', then 'it is at Tuesday' would mean 3rd P, while 'it is Tuesday' would mean 4th P.

24

Ultra Hal 7.0 / Re: problem questions

« on: August 14, 2012, 05:31:31 pm »

Yo, Yo
  To muddy-up 'why', I should have included:
1.2.  synon 'so'  ex: He needed eggs, so he went to the store.
   {'so' introduces a caused event, whereas 'because' introduces a causal event.}
3.2. ex: He needed eggs and went to the store.
   {causal events typically precede caused events.} 
   My earlier example: 'He went to the store and bought eggs.'  doesn't show
   any causality, but the temporal nearness hints at a possible infinitive: 'to buy eggs'.

25

Ultra Hal 7.0 / Re: problem questions

« on: August 07, 2012, 02:07:29 pm »

I remember 3 things I do to answer 'why':  ex: Why did he go to the store?
Looking at stored knowledge:
1. (explicit) synon 'because'  ex: He went to the store because he needed eggs.
2. (explicit) infinitives  ex: He went to the store to buy eggs.
3. (implicit) causality is temporal  ex: He went to the store and bought eggs.
   Of course, not all events are causal.  ex: He went to the store and saw his friend.
   'To buy eggs' is a stronger answer because of the inference chain:
      To go to a store is to shop.
      To shop is maybe to buy.
   

26

Ultra Hal 7.0 / NewB questions

« on: July 25, 2012, 06:02:17 pm »

  My first statement to HAL was: Today is Tuesday.
Of course HAL already knew the current date, but that's not how he read me.
He used his scheduling rules to remind me on the following Tuesday that that day was 'today'!
  Although HAL is good at chatting, that hides much of his true understanding.
Is there a brain with all the clever responses removed?  Does it do much other than echo the
dictionary? 
  As a chat-bot, evolved from ELIZA, HAL stores facts as narratives.  My own experiments
break the input into morphemes, and store facts as relations under each subject.  This
removes much of the repetition of determiners and adjectives from memory.
I suppose that inference can be made to work by matching keywords, but I find the morpheme/
sememe storage conducive to realistic understanding, for answering 'what happened', 'what do
you know', 'what did you do', etc.
  Thanks for listening; I am impressed by HAL but want a tool for exploring the algebra of grammar,
which has hidden aspects like trust, actuality, physicality, sequence...  Can HAL differentiate the
temporal 'it' from the non-temporal 'it'?