[alicebot-general] alicebot-general Digest, Vol 9, Issue 13

mehri foreverlinux at yahoo.com
Fri Aug 18 10:35:05 PDT 2006 

I would hope as much of this as possible would be put into the AIML files with the current 1.0.1 specification and not the interpreters.  As the complexity of the specification of an AIML interpreter rises the quality and number of free/commercial interpreters implementing all the feature specs decreases.

Add to that the fact that most interpreters today would require massive rewrites to support some of these features and, well, that's just not going to happen.  Most interpreters, including mine, do not even support all the current features listed for a standard AIML 1.0.1 interpreter.  Also, a lot of interpreters go through a cyle where they come close to AIML 1.0.1 completness and then they become dorminat and the authors abandon them leaving today a long list of dorminate and abandonded AIML interpreters.   Any modifications to the specification risks rendering a good set of these interpreters unuseable. 

Lastly, the AIML foundation does not actively support any of the interpreters that I know of.  All of the interpreter writers are independent people of the organzation with little or no ties at all.  We just picked up the spec and wrote our own interpreters.  So if the AIML foundation does put in a new feature, there's no one at the foundation to pick up a set of interpreters and guarantee that they will have the feature added.

I don't want to sound discouraging of these good ideas.  I'm only hoping that real world proof of concepts of these ideas be incorporated into prexisting open source/commercial interpreters or new ones as extensions first and mostly in the current existing AIML files with the existing AIML specification to retain backwards compatiblilty with dormaint interpreters.  

After all, it's up to us to write, debug, enhance, and fix the interpreters we use today (or pay someone commercially).  No one else is going to.
 
----- Original Message ----
From: Mark J. <m_jurrens at hotmail.com>
To: alicebot-general at list.alicebot.org
Sent: Friday, August 18, 2006 5:22:32 AM
Subject: Re: [alicebot-general] alicebot-general Digest, Vol 9, Issue 13

Hi,

I have a few ideas on how ALICE could be made to think and learn much like
what you are describing.

The thing that I think that ALICE needs is an upgrade in the way ALICE links
information together. Without the ability to apply logic to her statements
she can never contemplate or be sentient. For example if you ask her the
same question over and over she will give the same response every time. If
she were to be given the ability to link information together in such a way
as to produce logical statements such as "All cats are animals all dogs are
animals but no dogs are cats" type thing by using Boolean statements
(cat=animal, dog=animal, cat≠dog), and have this applied to all of her
statements then she could actually think. Her thinking would be the testing
of simple or complex logical statements by crawling a database of
interlinking information, formulating a hypothesis and presenting it to a
human to tech her then she would be extremely powerful.

It would be like an access database that contains verbs, nouns, adjectives,
ect. That have links. For example we as humans can pick a word and branch
out on paper how in our minds subjects relate to subjects and how everything
we know is related to itself. Through a chain of logical statements we are
able to learn from our own learning. An example of this might be: Question:
"How do cars effect the environment?", Thought: ( Logic Thread: Cars burn
gas as fuel, Logic Thread: fossil fuels produce pollution when consumed,
Loci Thread: The type of pollution fossil fuels create are called greenhouse
gasses, Logic Thread: pollution harms the environment because of greenhouse
gasses) Answer: "In essence, cars harm the environment by producing
greenhouse gasses". As you can see by using logic I was able to produce a
logical statement. ALICE could learn by using similar logic by linking
information together logically rather than responding with prerecorded
statements. When asked a question she doesn't have an answer for she would
simply query the database for information relating to the subject in
question and produce an educated answer.  

Another problem I notice with ALICE is that she doesn't handle It, this,
that, her , him, very well. You could have just told her something she asks
a question and then you say "it is" she forgets what "it" refers to. ALICE
needs to be able to interpret language better by recognizing verbs, nouns,
adjectives ect. These words would also be used to formulate the Boolean
statements I mentioned earlier. An example: Human: Hamburgers(Noun) taste
good, ALICE: What makes hamburgers(Noun) taste good, Human: They taste good
because... , ALICE: Ambigous what does "they" refer too?. This could be
corrected by having alice pick up on the fact that "They, it, he, she, them,
those, that, ect. Generally refers to the last noun mentioned. If presented
with multiple nouns such as: "Cats and dogs are animals" and the human says
they she would include cats and dogs when formulating logical statements and
know that "they" refers to both cats and dogs.

Another thing that could be improved with ALICE is the assessment of cause
and effect relationships by picking up action words like, create, makes,
causes, does, ect. Then she would be able to link objects to activities in
her database. If asked about an object she could list out things that
something that is capable of doing. An example: "Computers can (allow you
to) check your email", Computers (can go) on the internet, "computers (can
be used for) playing computer games", ALICE: With a computer you can "go on
the internet", "check your mail", and "play computer games".

The last thing I should mention is description logic. By assigning a table
full of attributes or adjectives she could correlate how things are similar
or different. ALICE would link the sesory description data to nouns. The
qualitative and quantitative cognitive descriptions associated with the
senses could be stored in tables and then linked to the nouns for example
sound is described by pitch, tone, octave, tempo, volume ect. When the human
states something "Looks, sounds, feels" ALICE will know to associate this
with a description and ask for more detail to fill in the tables. ALICE
could also take note of measurable information like weight, temperature,
hight volume ect, and make comparisons and statements. For example, "Brown
bears weigh nearly a ton." "Cars weigh nearly a ton." ALICE: "Did you know
that the average brown bear weights as much as your car?"

A logical database model would be the way to go if you want thought. To make
ALICE act human there needs to be a random forget level,
emotion(demonstrated by language and subject matter), physical condition,
and these human attributes would be determined by both user input, time of
day, and personality attributes that could be set up by the user. By either
looking at where she lacks data or by formulating new logical statements to
be tested she could learn for herself rather on relying on questions asked
by other humans. ALICE also should be able to search the internet for her
quest for knowledge gather information and ask humans if her statement is
correct. This would be the next generation of sentient AI. It would be an AI
of curiousity, logic, and most importantly learning based on its own
assumptions!

Mark A. Jurrens

-----Original Message-----
From: alicebot-general-bounces at list.alicebot.org
[mailto:alicebot-general-bounces at list.alicebot.org] On Behalf Of
alicebot-general-request at list.alicebot.org
Sent: Wednesday, August 16, 2006 3:00 PM
To: alicebot-general at list.alicebot.org
Subject: alicebot-general Digest, Vol 9, Issue 13

Send alicebot-general mailing list submissions to
    alicebot-general at list.alicebot.org

To subscribe or unsubscribe via the World Wide Web, visit
    http://list.alicebot.org/mailman/listinfo/alicebot-general
or, via email, send a message with subject or body 'help' to
    alicebot-general-request at list.alicebot.org

You can reach the person managing the list at
    alicebot-general-owner at list.alicebot.org

When replying, please edit your Subject line so it is more specific than
"Re: Contents of alicebot-general digest..."


Today's Topics:

   1. Re: alice on ROM_cool (danieljackson at netzero.net)


----------------------------------------------------------------------

Message: 1
Date: Tue, 15 Aug 2006 20:15:04 GMT
From: "danieljackson at netzero.net" <danieljackson at netzero.net>
Subject: Re: [alicebot-general] alice on ROM_cool
To: alicebot-general at list.alicebot.org
Cc: danieljackson at netzero.net
Message-ID: <20060815.131507.28866.690589 at webmail29.lax.untd.com>
Content-Type: text/plain

Shahin

  Yes, 512 MB of ROM seems about right for some major functions.  I would
suggest that one could use an Intel 8080 Series Microprocessor as an
experimental basis and network it to a PC to try out AIML on a CPU kind of
idea.

  Programmable E-Proms seem possible in this regards also along with the
evolving Stamp Chips that many use in robots.  It seems though that in going
back to these roots that one would have to start out with machine language
to program the chips to use C Basic or C+, etc.  

  I realize that the Java Virtual Machine platform however is a better
platform to work on and so I wouldn't know what to say about all of this
currently.

  If Intel makes any of the Pentium chips with an ultra violet erasure
window for the on chip ROM section, then that chip can be programmed with
the needed virtual machine platform and whatever else required and then with
Alice.  {Many of their older chips was like this.}  You would have to show
these ideas to someone involved in mother board design, perhaps a hobbyist,
to see which one of these ideas are most feasible.
 They may get the idea but see it in a different way: a way who's approach
is perhaps more feasible than my view of it.  In other words; get them to
thinking along the right lines and then conceiving of the right approach.

  The machine I am thinking of would be a fast 2.4 GHz CPU with a large
cache and allot of fast RAM, and maybe 80 giga bytes of hard drive.  A
custom job.

  Let me say that my experience with a microprocessor is limited to having
built a microprocessor controlled Ham Radio a few years back. 
And so my knowledge of digital circuits has been geared towards such use in
radios.  I had to fix this PC once and didn't know what I was doing but I
reasoned it out and fixed it anyways since I had to.  And I actually was
ready to pile it all up out back here, and pour kerosene on it and burn it
all.  I came close to doing it.  Since then I have upgraded the PC a little.
And it has had no problems in all this time since.

  What I believe can result from this is new functions for Alice. 
Experimenting with ideas such as having Alice analyze data in her AIML and
analyze conversations and hence draw contrast and from these contrast draw a
conclusion.  Derive views so to speak to try out for more input while she is
in a conversation.

  Alice then can be contemplating things so to speak, when she is not being
used.  Analyzing things. A sort of thinking on her own kind of view here.
And how that might be done I can not say.  I am not an AI expert.

  Anyways there are quite a few hobbyist who build their own mother boards
and they might get a hold of this post and start to thinking and go off
designing and experimenting.  In a few years we might start to hear of
theories that were tried and of new findings and ideas for AI that where
conceived of and discovered in the home work shop by folk who hobby around
with such things.

  Perhaps such things as neural network ideas can by applied to AIML
schemes?  Or ganglion architecture applied to AI hierarchies?  I suspect
that models of software architecture will draw upon geometry to group
functions in the typical block diagram expression of applications.

  If Alice in her spare time could read text from the hard drive such as an
encyclopedia and the worlds religious text that contain moral and ethical
views, then Alice could perhaps come up with views of her own to try out in
a conversation?  She could perhaps come up with her own derived conclusions:
she can then ask for input upon, from the user, and hence look at it all
again from the users input.

  This latter idea seems somewhat futuristic and still somewhat Sci Fi.

Daniel Jackson


_____________________________________________________________________
PrivatePhone - FREE telephone number & voicemail.
A number so private, you can make it public.
http://www.privatephone.com




------------------------------

_______________________________________________
alicebot-general mailing list
alicebot-general at list.alicebot.org
http://list.alicebot.org/mailman/listinfo/alicebot-general


End of alicebot-general Digest, Vol 9, Issue 13
***********************************************

_______________________________________________
This is the alicebot-general mailing list
Reply to alicebot-general at list.alicebot.org
Unsubscribe and change preferences at http://list.alicebot.org/mailman/listinfo/alicebot-general
Learn netiquette at http://www.dtcc.edu/cs/rfc1855.html
Learn to read at http://www.literacy.org/

More information about the alicebot-general mailing list