* This site is based on internet standards not supported by IE 6.
wJones
EC Container 5

Stored Purpose - Agency

Stored Purpose - a new approach to computing leveraging the current capability of information science

Subscribe

<< What is Philosophy of the Mind? | Stored Purpose Safety and Curtailment >>

Citation

Warren Jones, Lana Rubalsky (2010) "Stored Purpose - Agency", wJones Research, October 8, 2010
A better name than existential entities was needed for the workers of a metacomputer. We replaced the word person with existential entity in the definition below, and it fit the requirement precisely:

a·gent Noun /ˈājənt/
  1. An existential entity who acts on behalf of another, in particular
  2. An existential entity who manages business, financial, or contractual matters for an actor, performer, or writer
  3. An existential entity or company that provides a particular service, typically one that involves organizing transactions between two other parties

Figure - Components of Agent Identity<br />The Purpose of the Metacomputer Fabric is distributed as distinct agent Identities.  The Goals of each agent is regulated by the systems balancing equation.

Figure - Components of Agent Identity
The Purpose of the Metacomputer Fabric is distributed as distinct agent Identities. The Goals of each agent is regulated by the systems balancing equation.
We defined seventeen core agent classes to perform essential service and support functions for the metacomputer and a patron organization. The most important agent types are the nexus and assistant classes.
  • Nexus agents will assist the human ecopoesis leader in balancing system Identity and pursing the regulation Goal and it’s subForms (conservation, creation and learning). Nexus classes will serve as the senior-most member of the agency.
  • Assistant agents will serve people, learning the specific resources and responsibilities of their “client” and performing services on their behalf, much like a knowledgeable human assistant would. Assistants will use metacomputer and mobile device sensors (i.e. bluetooth, GPS, orientation) to track context and predictively align technology states with the needs of the client. The resultant semantic linkage between agent and person will resemble a wireless extension of Mica cellular and cortical communication. An agent’s ability to predictively employ technology based upon client context may give physically impaired users, the feeling that the agent-metacomputer has become an extension of their body.
To exist, an agent must be instantiated, by loading a read-only definition of their “class” into a metacomputer “database” called a soma. This read-only definition is called genetic Identity and will be the Construction part of their existence.
Figure - Resource and Responsibility Graph<br />Agent template design starts with the creation of a simple latticework of symbol relationships, free of Form data.  This data is prioritized and force directed, which logically draws highest priority and highest connection symbols toward the center.  This geographic layout becomes useful once the agent mind is instantiated as future related thoughts become metaphysically close.
Figure - Resource and Responsibility Graph
Agent template design starts with the creation of a simple latticework of symbol relationships, free of Form data. This data is prioritized and force directed, which logically draws highest priority and highest connection symbols toward the center. This geographic layout becomes useful once the agent mind is instantiated as future related thoughts become metaphysically close.
Building an agent template starts with a latticework of symbols linking responsibilities and resources. These symbols must cover Purpose required to sustain Self and to serve a client. That latticework is then augmented with Platonic Forms that describe universals, general symbols that can be downloaded from a global library. For a lawn maintenance bot, symbols would include responsibilities such as company, pedestrian and grass and resources such as water, basic earth, moon and celestial mechanics, and metacomputer instrument capability classes.

Logically each agent will have its own unique Identity, yet metaphysically, the sum of all Identities, i.e. the loci of all symbols, objects and logic, will be the Purpose of the greater metacomputer. The reason is is a result of the “balancing equation” that regulates all Purpose added to the metacomputer. A result of this design is that agents will suffer opposing Purpose for each Goal assigned.
  • This may be manifested within a single agent, such as a lawn maintenance bot with Purpose to both promote growth and cut grass.
  • It may also be distributed, such that a medication delivery bot may make tread marks on a floor each day, that another bot will have Purpose to clean.
Finally, some Purpose, such as the such as that of ecosystem rain, may be added to the metacomputer, and thus part of the balancing equation, yet not a Goal assigned any agent.

Agency

A person or organization’s Agency is somewhat analogous to the Applications directory of stored-program based computers. Each agent fulfills purpose serving one or more clients which may be a person (assistant class), an organization (nexus class), a class of instruments (instrument class) or one or more agent or human entities (all other agent classes). Agents each process Purpose for themselves and for supported clients. It is important to note that even though an agent may have a primary responsibility to process and support client Goals, agent Fabric and historical memory is always relative to the agent’s self and as such, different from the Fabric of the client.

The following is a brief summary of the standard agent classes that would be employed in a standard organization.

Agent Classes

assistant

personal assistance
processes and supports Purpose of supported client
interfaces with other agents on behalf of client
lifelong assistant
Categories
supervisor, full Stage

nexus

network or exoskeleton coordinator
manages exoskeleton and network kernels
manages core code stack (i.e. messaging, communications)
manages processing resource allocation
wakes and sleeps agents
approves architect activities
processes statistician experiments
processes enforcement requests

architect

optimizes optimizer, approves technology learning
optimizes interpreted purpose and goal logic
programs goal and schedule modifications
compiles logic byte code
compiles purpose and goals
prepares messenger byte code payloads
approves statistician requests for anomaly investigations (randomized tests to gather details regarding statistical anomalies)

messenger

transports information (names, data, security codes, byte code) between agents

stage

presents information
monitors availability of network assets
builds file system
builds pipes
manages foreign program interfaces

storekeeper

manages costs, constraints and allocation of shared resources
grants resource access requests
tracks resource availability
reports resource limitations to nexus
tracks costs and prepares billing information

planner

manages plans
manages plan memory (pre-memory)
processes plan optimization
processes prediction failure
triggers emotion response
manages pre-memory
manages planning kernels and research allocation
pushes action context
processes context value

librarian

manages symbolic knowledge
manages symbol representations
researches and gathers new symbols, relationships and routes
processes new symbol data for system resources
mediates data linkages with symbolic data sources (i.e. WordNet, DBPedia, local ontologies)

archivist

manages metacomputer information storage
manages physical - relative fabric memory
manages information storage and media type tagging
monitors recoverability of archival data
determines efficient organization methods for data storage and cataloging
prepares forward and reverse data propagation (core data needed by instrument agents to make local decisions for assigned purpose)

instrument

manages instrument resources and capability execution
executes capability algorithms
resolves resource constraints
manages local plan memory
notifies non-local network client planners of plan differentials

enforcer (conscience)

manages curtailment violation and purpose misalignment
monitors plan for purpose alignment failures
processes curtailment violation notices
submits plan changes to client agents
notifies client network planners of enforcement failures

reaper

recovers resources that no longer serve a purpose
discontinues agents
clears expired stream data
processes network resource recovery for client agents
returns un-met responsibilities to the architect
gathers agent experience data and sends it to the librarian

statistician

prepares optimization information for client planners
notifies planner of changes to system norms and routes
interprets prediction failure
identifies orphaned purpose
performs regression analysis of data collected by librarians
identifies statistical anomalies
categorizes anomalies (i.e.
optimization hunch
component failure indicator
curtailment violation)
prepares hunch verification tests (randomization variances in actions to map new goals via regression testing)
reports anomalies to the nexus and architect
designs goal and schedule modifications

editor

provides drawing, writing, symbol construction services to client agents via polymorphic canvas
provides user interface to edit fabric, media, route calculations, route data

sandman

manages agent quiescence for relative memory changes
puts agent to sleep - redirects real-time IO
wakes agent - enables real-time IO

keymaker

manages encryption services for client agents
generates public, private and one-time keys
replaces obsolete keys
processes security related curtailment violations
optimizes security processes

coreboot

first agent to load from host kernel, initializes agency upon cold start
starts nexus

Agent Categories

supervisor vs. subordinate

Supervisors must process the memory and thus adapt the Fabric of subordinate agents. Subordinate agents cannot process goals or routes that cannot be processed by supervisors. Subordinate agents do not possess any degrees of freedom greater than subordinate agents.

front Stage

Primarily performs physical world services

right Stage

Primarily performs symbolic services

left Stage

Primarily performs quantitative services

back Stage

Primarily performs virtual world services

seated vs. not-seated

Fixed position on stage or Roaming

on Stage vs. off Stage

Member of Network or not

locked Stack

Some or all goal pursuit logic is stored in read-only media. Device cannot optimize all facets of Context
partial lock
Will vary but typically means unable to change goals, add routes but can optimize route selection costs
full lock
Can select optimal routes but cannot update route costs and thus is unable to vary responses … not intelligent.

shopper (with credit)

Can engage off Stage entities in conversation and negotiation, can acquire resources

on Archive vs. in Stream

Memory goes to Archive (permanent) or to Stream (short term then purged)

<< What is Philosophy of the Mind? | Stored Purpose Safety and Curtailment >>

Related Documents


EC Container 6