don't let me confuse you with my terminology... use whatever terms you
like.
answers to questions inserted below
On 09/28/2014 11:04 PM, Piaget Modeler via AGI wrote:
It 's simply a matter of nomenclature you use the former, I use the
latter.
Recipe = Solution
Opportunity = Intention
Here are the definitions in my language:
(prototype Solution :Before :Steps :After)
(prototype Intention :Goal :Status :Depth :Urgency:Priority)
What are the attributes of an "Opportunity"?
Think of Opportunity as a record in a database. There are many fields
such as
name
benefit type
beneficiary
readiness
intrinsic worth
abstracted current value
version of evaluator
conditions for promotion
recipe name
recipe reliability
recipe time cost
recipe material cost
recommended by
confidence in recipe
(prototype Recipie :Preconditions :Steps :Effects)
(prototype Opportunity :Priority :Risk :Cost :Skills :Viability
:Elements)
(prototype StansSystem :CurrentRecipe :CurrentStep)
How does the Opportunity's :Priority get assigned?
The opportunity is promoted by a promoting routine. It is based on
factors given in the record of the opportunity and in working memory
(situational memory.) Every opportunity uses the same promoter
routine. This promoting occurs in parallel given optimised hardware.
Promoters produce a number - the higher number wins the promotion step.
Once the promoters are done, the highest ranked opportunity is
considered as a possible replacement to the currently selected
opportunity. There is a final arbitrator that must look at costs
associated with interrupting the current opportunity - restart, clean up
and so forth. If the promoted opportunity is clearly better than the
currently "running" opportunity, a switch of opportunity occurs. Given
that the opportunity was "selected," it now has top priority and it's
recipe will run until conditions cause a better opportunity to be
selected, or the opportunity is consummated - benefit delivered.
Is there a current recipe somewhere where we need to maintain a
pointer to the current step?
Program counter? The recipe will be running and it will be like any
other program that is written for execution. Perhaps the question is
about an executive type function?
An executive function will be running at all times either as another
thread or even another processor or separate computer. The executive
watches the clock and "progress" indicators. It coordinates the cycle
of promotions, selections and aborting the current opportunity. (there
are a few other parts to the system.)
Also, the better system would have promoters running and monitoring
changes in situation variables that could lead to "different" evaluation
of opportunity. (if no data changes, the promoter routine's result would
be the same.)
There is a management cost associated with the flexibility of being a
"general purpose" unit. One strength of the system is in the simple
cycle for making a new choice. The difficult parts of the system are in
the "mining" of opportunity, evaluations and promotions. Any
intelligent system will need to do these "difficult" things.
I like the way that a challenge becomes a new "opportunity" and as the
opportunity matures, it floats to the top and participates in the
working system.
<end of response>
Kindly advise.
~PM
------------------------------------------------------------------------
Date: Sun, 28 Sep 2014 22:38:38 -0600
From: [email protected]
To: [email protected]
Subject: Re: [agi] Are all goals created equal?
On 09/27/2014 10:17 PM, Piaget Modeler via AGI wrote:
...
But why would one goal get a higher priority than another, aside from
inheriting its priority from a basic need?
Perhaps I lack imagination.
Your thoughts?
~PM
Priority is part of a bigger picture. Priority can be “gen'd” from
many sources. Is there a way to manage the barrage of priority claims?
What priority will rise to the top and become the one we react to in
the next moment?
We want to discern between priorities because it is likely that a few
will indeed be “better,” that is, produce better outcomes in the
future. Indeed, this is a hard problem, and possibly resource
intensive. For us who ponder "how intelligent" a machine can be, it's
part of the challenge.
The original question relates to how goals get priority levels. My
description of how I think this can work follows:
In my “master plan” for an intelligent unit, priority is based on the
“recipe” that is currently being cooked. Recipes have steps, and the
“next step” is determined by what the recipe prescribes. This sounds
too simple to be useful, but there is a bigger picture that I hope
will explain why this “source” of priority can be effective.
As programmers, we know that things are accomplished by a sequence of
steps. Even in object programming, the methods are steps, that is,
sequential programming. If I start down a sequence of steps, then it
only makes sense that I follow through the routine. I started the
sequence because I want the affect that those steps will generate. Can
we agree at this point that if we are in a sequence, then the next
step is the priority?
Then, our “intelligence task” is to determine the sequence that we
should be executing at the moment. As hinted above, we have reasons
for starting a sequence and it is that set of reasons that form a
selection criteria. “Reasons” are built up in layers – things build
upon related concepts. As Jim Bromer has suggested, concepts can
quickly become complex and confusing. My plan attempts to control this
complexity by not attempting “logical gymnastics” with all these
concepts. Instead of attempting to be a black belt in logic, I want to
find other ways to select. Not so far fetched given that children
acquire methods before they have well developed logical grasp on the
concepts of life.
My approach is to look at recipes as being a component of a more
encompassing object, the opportunity. Opportunity is much more that
just a recipe, it also factors in risk, cost, skills, viability and
elements of the surroundings.
To simplify where I'm going with this, our intelligence task is one of
acquiring and comparing opportunity. Comparing is where we build the
basis for selecting the better opportunity for the moment.
Based on this strategy, I see “priority” as determined by the working
set of opportunity that the unit has adopted, or been given, or
otherwise acquired. Select the opportunity that fits the moment, work
the recipe of that opportunity and you are doing “high priority,” as
far as your unit is concerned.
It may seem that I've only shifted the task of selection of priority
from “the goal” to “the opportunity,” but as one attempts to determine
why his goal is a goal, and what value it has, he or she will soon be
faced with an “opportunity” like scenario. Just skip the goal thing
and go directly to opportunity, and allow that to be your basic currency.
Lots of details left out (long enough post.) I believe there is an
architecture here. Nearly all aspects of the system can be implemented
as “opportunity.” Building everything out of opportunity is an
interesting “bootstrapping” exercise. Result is a simple “machine.”
Stan
*AGI* | Archives <https://www.listbox.com/member/archive/303/=now>
<https://www.listbox.com/member/archive/rss/303/19999924-4a978ccc> |
Modify <https://www.listbox.com/member/?&;> Your Subscription [Powered
by Listbox] <http://www.listbox.com>
*AGI* | Archives <https://www.listbox.com/member/archive/303/=now>
<https://www.listbox.com/member/archive/rss/303/9320387-ea529a81> |
Modify
<https://www.listbox.com/member/?&;>
Your Subscription [Powered by Listbox] <http://www.listbox.com>
-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com
