PERSONAL EMERGENCY MANAGEMENT using the FREE LIFE PLANNER By Andrew John Dougherty --- Formalized Research Database: Cluster, Study and Apply (FRDCSA) Project --- Ontolog Summit 2022 Track 3: Environment Disasters April 20th, 2022 --- This talk is in memory of: My beloved sister Ann, whose tragic death when I was 10 years old drives to this day my unswerving commitment to the completion of this technology. Christina Taft's beloved mother, Victoria Taft, whose tragic death in the Paradise, California wildfire in 2018 drives her tireless commitment to emergency management technology. --- ## SECTION Introduction === --- 'Predicting, Monitoring, Mitigating and Managing' (Environmental) Disasters === This talk concerns 'predicting, monitoring, mitigating and managing' (environmental) disasters. --- 
Example Environmental Disasters:
 
  • Climate change
  • Environmental pollution
  • Wildfires
  • Floods
  • Habitat loss
--- I suggest what I (dryly) term a 'draconian solution' (i.e. reducing the problem to a harder problem). --- As an analogy, I 'reduce' the environmental disaster problem to general life planning, and put forth a satisficing implementation called the Free Life Planner aka FLP as a solution, --- and demonstrate how it applies in the specific case of handling environmental disasters. --- Preview of the Free Life Planner (FLP) --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/images/flp-screenshots-1.png === Here are some (outdated) screenshots of services that the Free Life Planner provides. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/images/flp-screenshots-2.png === This talk is going to focus on this intelligent personal assistant which will help individuals to take control of every aspect of their lives, including but not limited to emergency preparedness. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/clips/iem-20220327-cycle-weekly3.mp4#t=40,50 === This video shows one temporal planning subsystem called Interactive Execution Monitor version 2 (a.k.a. IEM2), which is part of the backend behind FLP. There is a corresponding web front end to this planning system. --- ## SECTION The Problem of Environmental Disasters === --- 
Some dimensions of the problem:
 
  • Prevention
    • Reducing individual contribution
      • E.g. committing to recycling, lowering
        electrical usage, etc
    • Reducing collective contribution
      • E.g. preventing industrial accidents
  • Management
    • Protecting people from disasters
      • Notification and evacuation
  • Sociological implications
    • Increased exposure of the poor
=== There are many dimensions to the problem of environmental disasters. --- In July of 2021 I wrote Al Gore regarding FLP, describing how one use-case dealt with helping people to reduce their environmental footprint. --- The use-case could be implemented via a particular kind of intelligent tutoring system, which elicits from the user all the relevant behaviors they engage in that affect their environmental footprint. --- FLP would then check whether or not these behaviors conform to (local) best practices in its rulebase. When they do not, the system would attempt to act as an accountability partner to help keep them on track towards modifying their behaviors. --- ## https://en.wikipedia.org/wiki/2004_Indian_Ocean_earthquake_and_tsunami/ file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/images/tsunami.png === A second example of an environmental disaster that could have been mitigated is the 2004 Indian Ocean earthquake and tsunami, which took over 200,000 lives. A cellphone warning system could have alerted and directed many more people to safety. This tragedy was disheartening because, although an idea is not an implementation, at that time I had been campaigning 2 years with no discernable progress for this system. 22 years on, although it hasn't yet made its impact, much technical progress on FLP fortunately has been made. --- https://frdcsa.org/~andrewdo/writings/artificialintelligence-second-half.txt === In 2003, I wrote an article on general purpose accident avoidance using Cyc and sensor networks. This article talked about using Cyc and KBS systems to make life-saving inferences. Although it describes a counterfactual scenario where a child using this system avoids being hit by a train, it bears strong resemblance to the problem of sending messages telling people within certain spatial regions to alter their behavior, such as to help move themselves, others nearby and loved ones to higher ground to avoid a tsunami. --- https://en.wikipedia.org/wiki/2021_European_floods/ === Or, moving to higher ground to avoid flash floods. The July 2021 floods in many European countries including Germany were an illustration of unprecedented weather saturating the emergency response coordination. --- These are a few examples of where FLP would be useful. However, being a life planner, it is intended to adequately manage many other situations and types of events. --- 
Some General Use Cases for FLP:
 
  • Natural and Man-made Disasters
    • Pandemics
    • Climate Change
  • Conditions
    • Poverty
    • Homelessness
    • Disability
    • Illness
    • Domestic violence
    • etc
=== Here are some of the use cases for FLP. --- ## FIXME: Strengthen https://owntracks.org/booklet/ === For positional reasoning, FLP currently integrates with the Own Tracks system, part of FLP's Location Logic subsystem. OwnTracks tracks users' location via cellphone, and is free/libre software. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/images/flp-subsystems.png === The FLP GitHub pages list many of its current subsystems, which are in various states of completion. --- ## SECTION The Free Life Planner (FLP) === --- 
About:
 
  • Free/Libre Social Friendly A.I.
  • For day to day life
    • Generating and walking through plans
    • Behavior modification
  • For emergencies
    • Emergency preparedness
    • Emergency management
=== I will now focus on how to create 'Free-Libre Social Friendly Artificial Intelligence' in service of the problems of day to day living, including planning and behavior modification, and also emergencies, through emergency preparedness and emergency management. This can be achieved through systems integration and software applications of Knowledge-Based Systems. --- 
Purpose:
 
  • Need Intelligent Personal Assistant
  • Ensure 'Last Mile Delivery' of services
  • Provides public health and policy data
    • With fine-grained consent or
      secure multi-party computation
=== There is a need for an intelligent personal assistant which is with the user in the trenches, enabling the last mile delivery of salient guidance, and conversely uploading evidence for public health. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/clips/iem2-wsm.mp4#t=0,22 === FLP is intended to 'eventually always' meet this need. FLP is a mobile friendly web application, which plans preactively, proactively and reactively - at multiple levels of granularity - in order to both survive and thrive on a daily basis, and prepare for managing emergencies when necessary. --- 
Goals of the talk:
 
  • Convince the audience:
    • How zero marginal cost (ZMC) + libre + gratis
      properties enable ~8,000,000,000 potential FLP
      users
      • Precedent: GNU/Linux
    • That 'FLP could help save lives, improve quality
      of life and reduce suffering'
    • To enlist your support in completing,
      releasing and promoting FLP and FRDCSA
=== The purpose of the talk is threefold. First, to illustrate that because free/libre software is freely redistributable and can assist people indirectly, whether or not they have a computer, it has the potential to offer help to almost everyone on the planet, in the same way that GNU/Linux could. Second, to convince you that FLP could in fact help save lives, improve quality of life, and reduce intense suffering. Third, to enlist your active support of these projects. --- 
Disclaimers:
 
  • I am partially self-taught (so a lot of what I
    say here hasn't been independently verified)
  • I have difficulty summarizing
  • First video generated by my PresGen system
  • Apologize for the use of Text-To-Speech
=== Some disclaimers. --- 
Suggestions:
 
  • Please re-watch and occasionally pause
  • Please type any questions into the chat
    which I can then answer
=== And some suggestions. There is more relevant information than I am able to summarize contained in this presentation, so I recommend re-watching the video and occasionally pausing it as needed. --- 
Outline of talk:
 
  • Introduction
  • The Problem of Environmental Disasters
  • The Free Life Planner (FLP)
  • Plan-Monitor
  • Jason/AgentSpeak(L)
  • 'Predicting, Monitoring, Mitigating and Managing' Threats
  • The WOPR System
  • Prepping and Survivalism
  • The PrologCyc Discipline
  • E2C, NLU and NLU-MF and Rule Mining
  • A.I. Planning Technologies in FLP
  • Releasing FLP
  • The FRDCSA Project
  • Releasing the FRDCSA
  • FRDCSA Theory
=== Outline of the talk. --- ## https://www.youtube.com/watch?v=mk35eReQKvo file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/clips/Paradise-Camp-Fire-Evacuation---11-13-2020---v1-mk35eReQKvo.mkv#t=60,74 === This video by Paul Fierro shows his excellent simulation of a counterfactual evacuation of Paradise, California from the Camp Fire that started on November 8, 2018. One way to frame the problem of emergency management focuses on the 'global', 3rd-person, top-down, macro perspective. --- https://upload.wikimedia.org/wikipedia/commons/1/19/Woolsey_Fire_evacuation_from_Malibu_on_November_9%2C_2018.jpg === Another way to frame the problem is to focus on the "local," 1st-person, bottom-up, micro perspective. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/clips/plan-monitor-cell-phone-editted.mp4#t=0,15 === This can be achieved for instance via a mobile hands-free plan monitoring application, such as the one I developed shown here. --- https://live.staticflickr.com/4465/37560568716_61d7508169_b.jpg === I use a fisheye lens here as a metaphor for this 1st person perspective. FLP operates from this individual perspective, which may be federated to form a community, and thus focuses on the community perspective as well. --- "File:Woolsey Fire evacuation from Malibu on November 9, 2018.jpg" by Cyclonebiskit is marked with CC BY-SA 4.0. To view the terms, visit https://creativecommons.org/licenses/by-sa/4.0/?ref=openverse === --- "Canyon Fire 2" by /\ltus is marked with CC BY-NC-ND 2.0. To view the terms, visit https://creativecommons.org/licenses/by-nc-nd/2.0/?ref=openverse === --- ## SECTION Plan-Monitor === --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/behavior-tree-task-manager-for-covid-19-1-2.ppm.jpg === Originally, I was to give a presentation on the Pandemic Track regarding my COVID-19 Behavior Tree paper that recommended developing a Prolog-based 'plan monitoring system', but fell short of fully implementing one. The plan monitoring system was conceived to help overcome the last mile delivery of efficacious knowledge and procedures for reducing the spread of and mitigating the effect of COVID-19. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/behavior-tree-task-manager-for-covid-19-3-4.ppm.jpg === Unfortunately this paper, written at the very beginning of the pandemic, really missed the mark. I had not predicted that a sizable percentage of the population would not be willing to adhere to guidelines for measures like wearing masks or being vaccinated. Any successful disaster management strategy must therefore account for the human factors present. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/behavior-tree-task-manager-for-covid-19-5.ppm.jpg === Nevertheless, finishing the original Plan-Monitor would still be of specific use to those who need extra layers of security during the pandemic, for instance the immunocompromised and their associates. As well, Plan-Monitor is of general use for interactively walking users through plans. --- 
# As root
export URL='https://github.com/aindilis'
cpanm $URL/plan-monitor/raw/master/plan-monitor-0.01.tar.gz
cpanm $URL/perl-btpm/raw/master/perl-btpm-0.01.tar.gz
cpanm $URL/perl-btsk/blob/master/perl-btsk-0.01.tar.gz
=== So I persisted and ultimately developed a Perl-based Plan-Monitor system, in addition to the unfinished Prolog-based system, which can be installed using the commands shown, but which did not depend on any unreleased FRDCSA code. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/clips/plan-monitor-voice.webm#t=0,30 === Thus, Plan-Monitor is a mobile-friendly WebUI for walking users through tasks such as going to the doctor, the grocery, receiving mail, etc. Here it is in operation. Some important features shown here includes the ability to branch on plan failure and hands-free operation. --- 
Strengths of Plan-Monitor:
 
  • Based on behavior trees
  • Real-time (including WebSocket-based WebUI) 
  • Ability to handle nonlinear/branching/contingent
    plans
  • 'Imperative reactive planning-with-sensing, and
    execution'
  • Hands-free operation
=== Plan-Monitor has many strengths. --- 
Limitations of Plan-Monitor:
 
  • No concurrent tasks and plans
  • No deliberation 
  • Simple implementation
  • Currently lacking more sophisticated
    behavior tree features
=== However, Plan-Monitor also had some major limitations. --- After asking the A.I. planning community, there appear to be few options for an (open-source) system like Plan-Monitor that can handle temporal, concurrent plans. --- So I intend to fork plan-monitor and extend it to handle temporal concurrent plans via behavior trees, as a front-end, and handle the backend and merging of the concurrent plans using the previously mentioned Interactive Execution Monitor v2 (IEM2), more of which will be shown later. --- sanitize_surface(Person,Surface) ==>> ensure_state(wearing_gloves(Person,_Gloves)), true_state(in(Surface,Room)), true_code(isa(Room,room)), ensure_state(well_ventilated(Room)), true_state(holding(Person,HouseholdCleaningSprayOrWipe)), true_code(isa(HouseholdCleaningSprayOrWipe,cleaningSpray)), act(read_label(Person,labelFn(HouseholdCleaningSprayOrWipe))), act(use_as_directed_on(Person, HouseholdCleaningSprayOrWipe, Surface)). === Here is an example taken from my COVID-19 paper. However, the Prolog-based Behavior Tree software that evaluates this particular behavior tree has not yet been fully completed. --- ## SECTION Jason/AgentSpeak(L) === --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/images/jason.png === For the time being, I've taken a shortcut which will hopefully be backwards compatible. This shortcut is to switch to using the Jason Belief Desire Intention (BDI) Agent Architecture. --- 
+!sanitize_surface(Person,Surface) <-
        !wearing_gloves(Person,_Gloves);
        ?in(Surface,Room);
        ?isa(Room,room);
        !well_ventilated(Room);
        ?holding(Person,
                 HouseholdCleaningSprayOrWipe);
        ?isa(HouseholdCleaningSprayOrWipe,
             cleaningSpray);
        read_label(Person,
                   labelFn(HouseholdCleaningSprayOrWipe));
        use_as_directed_on(Person,
                           HouseholdCleaningSprayOrWipe,
                           Surface).
=== This is what the code from the earlier slide probably looks like in Jason/AgentSpeak. Fortunately, Jason/AgentSpeak appear to implement a superset of the features I was able to implement in the Behavior Tree software I wrote. It is also much more mature and popular. --- 
One possible mapping from our Behavior Tree software
to Jason/AgentSpeak(L):
 
  • !X ⇐ ensure_state(X).
  • ?X ⇐ true_state(X).
  • ?X ⇐ true_code(X).
  • X : .ground(X) ⇐ act(X).
  • ?x(X) : not .ground(X) ⇐ act(X).
  • +X ⇐ begin_state(X).
  • -X ⇐ end_state(X).
=== Here is one possible mapping from Behavior Trees to Jason/AgentSpeak. --- ## https://www.mdpi.com/electronics/electronics-10-02136/article_deploy/html/images/electronics-10-02136-g004.png file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/images/electronics-10-02136-g004.png === The default intention handling code for the Jason/AgentSpeak BDI architecture event loop operates in a round-robin fashion, and is thus FIFO workflow. This helps to avoid dropping the ball, as happens with a LIFO workflow. Because Jason is based on Prolog Interlingua, I have since integrated Jason with SWI-Prolog. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/incoming/jason-swipl-integration.png === I intend to use Jason/SWIPL as the metaplanner to control other (PDDL/etc) planners through FLP's planning wrappers. --- ## FIXME: Logicmoo needs a slide 
Metaplanning Capabilities:
 
  • Definitely will use the Logicmoo system when stable
  • Have not been able to get OpenCyc's planning and
    agency abilities to work
  • Planning on connecting Jason/SWIPL with OpenCyc
    through either Logicmoo's code, or perhaps using
    my own Cyc wrappers in Perl/Emacs-Lisp
=== Metaplanning could eventually be achieved through integrating Logicmoo, OpenCyc, Jason and SWIPL. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/Temporal-Planning-and-Inferencing-for-Personal-Task-Management-with-SPSE2-1-2.ppm.jpg === Another tool mentioned in my COVID-19 paper is the Shared Priority System Editor version 2, or SPSE2 system. This tool proved personally useful during the pandemic. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/images/spse2-pandemic-example.png === I often developed 50 to 100 point temporal plans for the accomplishment of tasks rendered more risky due to the pandemic. SPSE2 was a predecessor of, and now a component of, the Free Life Planner. --- ## SECTION 'Predicting, Monitoring, Mitigating and Managing' Threats === --- ## https://frdcsa.org/~andrewdo/predicting-monitoring-mitigating-and-managing-threats.pdf#page=1 ## file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/predicting-monitoring-mitigating-and-managing-threats-01.ppm.jpg file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/predicting-monitoring-mitigating-and-managing-threats-01-02.ppm.jpg === Because of a family emergency, I had to withdraw my Ontolog 2022 Pandemic Track talk. However, I had already written some of a paper which was targeted towards general threat management. --- ## https://frdcsa.org/~andrewdo/predicting-monitoring-mitigating-and-managing-threats.pdf#page=2 ## file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/predicting-monitoring-mitigating-and-managing-threats-02.ppm.jpg file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/predicting-monitoring-mitigating-and-managing-threats-03-04.ppm.jpg === I am now presenting in the Ontolog 2022 Environment Disasters track, but this is just as well, since the overarching tool this presentation is about - the Free Life Planner - subsumes both use cases. --- ## https://frdcsa.org/~andrewdo/predicting-monitoring-mitigating-and-managing-threats.pdf#page=3 ## file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/predicting-monitoring-mitigating-and-managing-threats-03.ppm.jpg ## file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/predicting-monitoring-mitigating-and-managing-threats-05-06.ppm.jpg file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/predicting-monitoring-mitigating-and-managing-threats-07-08.ppm.jpg === I wrote this paper, in order to talk more about threat management. --- ## https://frdcsa.org/~andrewdo/predicting-monitoring-mitigating-and-managing-threats.pdf#page=4 ## file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/predicting-monitoring-mitigating-and-managing-threats-04.ppm.jpg ## file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/predicting-monitoring-mitigating-and-managing-threats-07-08.ppm.jpg file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/predicting-monitoring-mitigating-and-managing-threats-09-10.ppm.jpg === This paper will be finished and released some time after the presentation. --- ## SECTION The WOPR System === --- "In a game-theoretic formulation of real world security protocols, it is possible to try ''to treat the world as a 'game' and then win that game by proving that bad things don't happen to people.''..." --- "...Yet, the cardinality of possible moves is so large that the branching-factor dominates, rendering (deeper) searches ineffectual (except in restricted subdomains)..." --- "...This suggests a knowledge-based approach towards inventorying and categorizing threats." - from the upcoming ''Predicting, Monitoring, Mitigating and Managing' Threats' paper --- file:///var/lib/myfrdcsa/codebases/minor/free-wopr/frdcsa/project.jpg === Our system for managing threats to security is called WOPR, after the contingent planning computer in the movie "War Games." --- 
Through WOPR and other systems, FLP is intended
to help bolster every type of security:
 
  • Social
  • Financial
  • Emotional
  • Food
  • Water
  • Cyber-Physical
  • Home
  • National
=== Through WOPR and other systems, FLP is intended to help bolster every type of security. --- %% Facts isa(andrewDougherty,person). %% Rules planForContingency(whatIf(infected(Person,covid19))) :- isa(Person,person). possibleResponse(whatIf(infected(Person,covid19)), quarantine(Person)) :- isa(Person,person). === So, if isa(andrewDougherty,person) were in the WOPR Prolog factbase, the Prolog code declares a contingency plan needs to be developed for the possibility that I became infected with covid19, and suggests that I should quarantine. The next step is to develop the hypothetical contingency plan for me quarantining. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/images/flux.png === In order to develop contingency plans such as the quarantine plan, we must deal with the reality that the life planning environment is dynamic, uncertain, incomplete, underspecified, etc. Thus, it is difficult to completely represent future states in which fluents like 'infected' might hold. One system which can handle such environments is the General FLUX cognitive agent programming system. --- 
Advantages of using General FLUX:
 
  • Under a free/libre license
  • Prolog-based
  • Can model state in the fluent calculus as a list,
    with an unbound variable at the tail signifying
    unknown fluents
  • Can make more progress now that I have more Prolog
    experience
=== FLUX has many advantages. We did implement about 10 different planning domains in FLUX. --- 
Difficulties using General FLUX:
 
  • Relatively slow
  • Apparently cannot handle nested fluents
  • Uses integers instead of atoms for arguments,
    which we overcame using the SerPro mapping
  • Have not yet received access to FLUXPlayer
  • Thus, no existing planning algorithms, so I 
    wrote Breadth-First Search
=== However, I had some difficulties when using FLUX for this purpose. --- 
Advantages of using Jason/AgentSpeak(L):
 
  • Designed for dynamic uncertain environments
  • BDI Agent Architecture
  • Based on Prolog, handles nested terms with ease
  • Has design patterns to implement other types of
    goals, such as maintenance goals
  • Peleus introduces limited STRIPS planning
    abilities (based on a reductive interpretation of
    Jason actions to preconditions and effects),
    possibly making Jason more deliberative
## https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.66.5917&rep=rep1&type=pdf === Because of these difficulties, I was unable to proceed until I rediscovered the Jason/AgentSpeak system, which is designed precisely to handle dynamic, uncertain environments. --- %% WOPR ?- planForContingency(X). X = whatIf(infected(andrewDougherty)). ... ?- possibleResponse(whatIf(infected(andrewDougherty)),Y). Y = quarantine(andrewDougherty). ... === WOPR is able to generate requests for the creation of plans for these contingencies, which are then implemented as plans inside Jason/AgentSpeak. --- 
%% Jason/AgentSpeak quarantine plan
+!quarantine(Person) <-
	?hasRoom(Person,Room);
	!at(Person,Room);
	+obligation(doNotLeave(Person,Room));
	...
=== And then the Jason/SWIPL plan is selected and executed. --- 
Open Questions with WOPR/Jason:
 
  • How do I handle time?  Sequentially? Temporally?
  • How do I model the environment? (EventCalc, SitCalc,
    etc)?
=== Some open questions - how do I handle time or model the environment? --- genls([impact,likelihood],risk). genls([vulnerability,threat,risk],item). genls([person,property,facility],target). argIsa(hasThreat/2,1,vulnerability). argIsa(hasThreat/2,2,threat). argIsa(hasRisk/2,1,threat). argIsa(hasRisk/2,2,risk). argIsa(isExposedToThreat/2,1,target). argIsa(isExposedToThreat/2,2,threat). ... === This is the beginning of a threat ontology for WOPR. Any threat knowledge-base that I could piece together for this talk would be incomplete, since I haven't done enough Natural Language Processing. --- %% Some facts genls(tornado,disasterEvent). genls(wildfire,disasterEvent). genls(tsunami,disasterEvent). genls(flooding,disasterEvent). genls(hurricane,disasterEvent). %% %% Prepare for all known disasters affecting all known persons planForContingency(whatIf(situationAffects(DisasterEvent,Person))) :- genls(DisasterEventType,disasterEvent), isa(DisasterEvent,DisasterEventType), isa(Person,person). %% Prepare for if any todo list item is not completed planForContingency(neg(inTime(completed(Task)))) :- task(TaskID,Task,Criticality), \+ completed(TaskID). %% Mark contingency as unprepared for if no known possible response unpreparedForContingency(Contingency) :- planForContingency(Contingency), \+ possibleResponse(Contingency,Response). === This is merely intended to be evocative of the kinds of threats that should be included, and how FLP will deliver on these counts. We could work on both preparations and appropriate responses ordered by some norm on likelihood and impact. --- ## SECTION Prepping and Survivalism === --- 
Goals of FRDCSA (and FLP):
 
  • The evasion of chance in survival
  • To minimize unnecessary psychological suffering
=== My A.I. has been from the outset intended to improve survivability while minimizing intense suffering. These are good goals when considering prepping and survivalism. --- 
Related Aspects of Free Life Planner:
 
  • Focus on preparations, they are easier to automate
  • Preparations can be likened to the first error
    message in an error log.
    • You don't debug the later messages, you debug the
      first, since the later might be a consequence of
      or otherwise affected by the first
  • Preparations are usually less time-critical
=== To partially achieve this, we rely on preparations. --- ## file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/EmergencyPreparednessChecklist.pdf file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/EmergencyPreparednessChecklist-1-2.ppm.jpg === These kinds of checklists are a start, but better is a persistent assistant which guides the user to achieve and stay on top of emergency preparedness. --- scheduleTask(flp, Person, checkBatteriesOnAllFireAlarmsInHome(Person, Home)) :- ought(maintain(Person,Home)), lastTime(happened(Time, checkBatteriesOnAllFireAlarmsInHome(Person, Home))), currentTime(CurrentTime), duration(Time,CurrentTime,months(Duration)), Duration > 6. === Here the system would schedule a task to remind the user to check their fire alarm batteries six months after the last time they checked. --- %% %% maintenance of victuals %% 1 months after the last inspection hasRecurrenceSpec(do(Person,inspect(Person,Victuals)), sinceLastTime(months(1)), [10:30:0,12:0:0,15:0:0]) :- ought(maintain(Person,Victuals)), isa(Victuals,victuals). === Here is a recurrence specification to remind the user to inspect their food and water supplies once a month, provided they are obligated to maintain them. --- https://frdcsa.org/~andrewdo/writings/homeless-story.html === This survivalism use-case story details how FLP is intended to work. It is definitely worth the read, many people including some Computer Scientists did not really understand FLP or FRDCSA until reading this document. --- https://frdcsa.org/~andrewdo/writings/health-story.html === This is another use-case story, detailing the healthcare component. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/images/free-life-planner-startup.png === I cannot in general show the full boot sequence of FLP because of private information, since I am dogfooding the application, i.e. using the program while writing it. --- ## SECTION The PrologCyc Discipline === --- possible(developsAilment(proteinPoisoning,Person)) :- isa(Person,person), isa(Meat,meat), hasProperty(Meat,lean). isa(Event1,eatingEvent), doneBy(Event1,Person) consumedObject(Event1,Meat), neg( holdsIn( timeIntervalInclusiveFn(now), and(doneBy(Event2,Person), isa(Event2,eatingEvent), consumedObject(Event2,Object), hasProperty(Object,fatty)))). === This is a good time to talk about FLP's usage of what I call the PrologCyc discipline. "PrologCyc discipline" means to me effectively following the protocols, APIs, vocabularies, etc that tools like OpenCyc have established. Here is a concept from the outdoor survivalism literature. What this rule is attempting to say is that "If a Person eats lean meat, and does not eat anything fatty along with it, it's possible that they will develop protein poisoning." The idea here is to use these kinds of rules as world knowledge to constrain plans. However, since translating this rule to PrologCyc, I have learned that this concept about eating lean meat being potentially harmful, is itself not necessarily correct. I will talk later about rule provenance and trust management. --- 
FLP's Knowledge-Based Systems Features:
 
  • The current FLP KBS API features are pretty minimal:
    • Highlights:
      • some poor isa/2, genls/2 support
      • allTermAssertions/1,2
      • constant_apropos/2
      • constant_complete/2
      • atTime([2022-02-19,21:16:29],Prop)
      • hasTruthValue/2
      • many others
    • Concerns:
      • incomplete support for and usage of
        microtheories in FLP right now
      • hence no multi-user support
      • no at-rest encryption of private data, etc
=== FLP has limited KBS support at present. --- 
The future FLP KBS API should be more impressive:
 
  • Talk via Pengines with Logicmoo
  • Use Logicmoo's Cyc access tools
  • Use Logicmoo's inference engines
  • Use FRDCSA's Cyc interface
  • Use FRDCSA's FreeKBS2 inference
    (Prolog and FOL)
  • Abduction of argIsa relations
  • etc, etc, etc
=== However, it will be much better once extended, and/or integrated with Logicmoo and or OpenCyc. --- ## SECTION E2C, NLU and NLU-MF and Rule Mining === --- https://frdcsa.org/~andrewdo/projects/E2C.txt === A major strength of FLP is that so-called 'rules' governing life can eventually be mined from natural language. --- ## SECTION A.I. Planning Technologies in FLP === --- ## file:///var/lib/myfrdcsa/codebases/minor/flp-advocacy/data-git/types-of-planning-1.txt ## === ## Some types of planning that I intend to cover with FLP. ## --- https://frdcsa.org/~andrewdo/writings/homeless-story/hygiene.p.pddl.LPG.sol.html === Here is a sample PDDL plan to tackle a homelessness scenario. --- ## file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/clips/iem2-3.mp4#t=100,120 ## --- 
PDDL has major limitations, such as:
 
  • Requiring perfect information
  • Inability to deal with uncertainty
  • Generally unable to sense or react outside of
    certain specialized planners
  • Inability to fail gracefully when no plan may
    be found
  • Inability to create new objects at run-time
  • Inability to modify the goal stack at run-time
  • Inability to have nonlinear/branching plans
  • Lack of inborn plan execution monitoring
=== PDDL is an extremely powerful planning language, which I use throughout FLP and FRDCSA, however it has some limitations. Please note There are probably exceptions to some of these cited limitations. --- 
Some Metaplanners:
 
  • Jason/SWIPL
  • PUCCINI
  • TAEB/Planner 
  • BotHack
  • Logicmoo’s adv_axiom.pl
  • FLUX-FRDCSA
  • Andrew Gree’s metaplanner
    • https://github.com/hyperc-ai/metaplanner
=== Since most PDDL planners to not seem to allow one to change the domain or problem specifications at run time (with the exception of Andrew Gree's metaplanner), we return to the subject of metaplanning. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/209_Performatives_to_engender_the_creation_of_performatives-1.ppm.jpg === In order to define metaplanning, I refer to this sample document written by Douglas Miles which describes what is intended by the term. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/clips/TAEB.mkv#t=15,38 === THE TAEB Planar AI makes heavy use of metaplanning and pathfinding for the video game called NetHack. There is for instance a higher than expected degree of similarity between playing this game well and surviving emergencies such as environmental disasters or homelessness. Both problems are roughly speaking survival-oriented pathfinding. I therefore rehabilitated the TAEB Planar AI in order to attempt to make use of its code for application to this real-life survival planning. --- https://frdcsa.org/~andrewdo/archive/taeb/taeb.html === This article goes into great depth about how TAEB/Planar's pathfinding works. It is very enlightening. However, I was not able to properly harness TAEB/Planar. I have had more success with metaplanning by focusing on Jason/SWIPL. ## BotHack: ## [Show clip of BotHack] ## Reliance on pathfinding ## Logicmoo’s adv_axiom.pl ## Andrew Gree’s metaplanner ## FLUX-FRDCSA --- Here are some of the contemplated use-cases for metaplanning with our autonomous Jason/SWIPL software. --- 
 • Help our software to carry out tasks
 • Copy itself onto the user's new computers
 • Help secure our networks
 • Remote control OSes
 • Security harden OSes
 • Help the user execute safe and advisable programs
 • Backup data and provide contingencies for it
 • Search all the files on the user's personal network
 • Handle sensor networks
=== --- 
 • Gather software
 • Package software
 • Scrub and redact software
=== --- 
 • Dialog with people about their situations
 • Monitor the health of people
 • Help people administer first aid
 • Help the impoverished survive
 • Plan for financial futures and contingencies
=== --- 
 • Help ensure self-discipline
 • Manage distractions
 • Stay productive
 • Help people manage their mood
 • Help walk us through challenging situations
 • Avoid emotional triggers
 • Help people respond to domestic violence
 • Develop prevention and response plans for
   various contingencies
 • Conduct thorough emergency preparedness and relief
   (including for disasters such as pandemics)
=== --- 
 • Browse the web and research for us
 • Crawl the internet and extract software and other
   artifacts
 • Process incoming downloads and forward chain with
   that information to accomplish more tasks
 • Carry out epistemic reasoning over data structures
=== --- A major goal of FLP is helping people avoid or get out of proximity to traps, since it is easier to get into a trap than it is to get out. --- ## Here is a formal definition of a trap, from the book Heuristic Search: ## --- ## ## "Many state space problems ... are reversible; that is, for each ## action a ∈ A there exists an action a−1 ∈ A, so that a(a−1(u)) = ## u and a−1(a(u)) = u. Each state that is reachable from the start ## state can itself reach the start state. ..." ## === ## "Many state space problems ... are reversible; that is, for each ## action a ∈ A there exists an action a−1 ∈ A, so that a(a−1(u)) = ## u and a−1(a(u)) = u. Each state that is reachable from the start ## state can itself reach the start state. ..." ## --- ## "... Hence, if the goal is reachable, then it is reachable from every ## state. Directed state space problems, however, can include ## dead-ends. ..." ## === ## "... Hence, if the goal is ## reachable, then it is reachable from every state. Directed state ## space problems, however, can include dead-ends. ..." ## --- ## Definition 1.10. (Dead-End) A state space problem has a dead-end u ∈ S ## if u is reachable and Pu = (S,A,u,T) is unsolvable." ## === ## Definition 1.10. (Dead-End) A state space problem has a dead-end u ∈ S ## if u is reachable and Pu = (S,A,u,T) is unsolvable." ## --- I want to make very clear that this project aims to help everyone, not just the marginalized. Having said that, there is a current focus on problems affecting the marginalized, since they are more exposed to more problems, such as the COVID-19 pandemic, and increasingly frequent and destructive natural disasters due to climate-change. Moreover, free/libre software inherently has more potential to reach the marginalized and adapt to their conditions in ways that proprietary software cannot. --- 
Other Important FLP/FRDCSA systems:
 
  • World State Monitor (WSM)
    • A digital twin
    • Has fluents editor, will learn which fluents
      are more volatile, and will thus more frequently
      assess their value
  • Interactive Execution Monitor v2 (IEM2)
    • Walks users through plans
  • Verber
    • Wrapper for various mostly PDDL planners
  • Many others
=== There are around 125 released FRDCSA systems on GitHub. Here are a few ones related to this presentation. --- 
Need to develop FLP2:
 
  • Need microtheory, multi-user, at-rest data
     encryption, HIPAA compliance, etc support
  • Need lazy loading of modules
  • No tests yet
  • Lots of spurious warnings and such that need
    to be fixed
  • Commenting is done incorrectly
  • Many other technical debts
=== We need to do a rewrite of FLP once Logicmoo is ready, as there are just too many issues with the present FLP codebase. --- ## SECTION Releasing FLP === --- 
FLP project success is defined as:
 
  • https://freelifeplanner.org
  • https://localhost:8443/flp
  • 1-800-GET-HELP
  • OS packages for FLP
    • `sudo apt-get install free-life-planner`
    • `sudo yum install free-life-planner`
    • `sudo pacman -S free-life-planner`
    • ...
=== Defining FLP project success. --- 
Issues with releasing FLP:
 
  • Inseparable from the FRDCSA system
  • FRDCSA hasn't been properly released
=== Releasing FLP (and FRDCSA) has been the Achilles heel of the project, because of difficulty redacting private information, writing the installer, learning the packaging mechanisms, hardware failures, and many other issues. --- ## SECTION The FRDCSA Project === --- https://altruisticsoftware.org/frdcsa/ === The 'Formalized Research Database: Cluster, Study and Apply' Project. --- file:///var/lib/myfrdcsa/codebases/releases/frdcsa-reference-0.1/frdcsa-reference-0.1/visual-aid/frdcsa-small.jpg === The project logo blends the F from FRDCSA, the satisfaction relation, and the Greek letter Phi (made by altering the Debian project logo). Phi here denotes a set of sentences representing problems or constraints. The FRDCSA Project logo is intended to suggest that the FRDCSA can correctly solve life problems, through extension of the Debian Universal Operating System with additional (usually AI) packages. --- file:///var/lib/myfrdcsa/codebases/minor/debconf-2020/video/resources/old/images/visual-aid.ppm.jpg === I want to create a increasingly powerful 'weak AI'. ## I am not even sure what all the acronyms mean these days (AGI/SAI/HLAI/etc). --- 
FRDCSA Ontology Projects
 
  • CO - Comprehensive Ontologies
  • CSO - Comprehensive Software Ontology
  • CLO - Comprehensive Library Ontology
  • CFO - Comprehensive Function Ontology
  • CPMO - Comprehensive Package Management Ontology
  • CFRO - Comprehensive FRDCSA Resource Ontology
=== There are many ontology and knowledge-base projects in some state of development in FRDCSA. --- 
More FRDCSA Ontology Projects
 
  • MKM project (Mathematical Knowledge Management) seeks
    to formalize mathematical knowledge within Cyc
  • Textbook Knowledge Formation: Really need dmiles'
    E2C program here
  • Research-Ontology 
  • Food-Ontology 
  • Reward-Ontology / Penalty-Ontology
  • ...
=== --- 
FRDCSA Project goals:
 
  • ~400,000 project todo list items, need to be
    vetted and redacted
  • Move beyond 1-person effort
  • Possibly get funding for bringing under test,
    redacting, releasing, additional development
  • Completely forkable dev ops, pros cons (con is
    fragmentation)
  • Finish getting a modification invariant UID for each
    do-todo-list-mode entry
  • Have people make packages or finish semiautomatic
    packaging using ESP(Jason<->SWIPL)
=== FRDCSA project is more fully specified by its many goals. --- 
More FRDCSA Project goals:
 
  • Easier to change circumstances than minds
  • Mass adoption of FLP
  • Extensibility and configurability
  • Mechanism design/gamification
  • Works for everyone but especially helpful to
    marginalized populations/demographics
=== --- 
Some actual mistakes the project has made:
 
  • Written too much bespoke software
  • Not enough packages of other people's software
    • Mainly due to difficulty building the Packager
      system
      • (Now working on Jason/SWIPL-based rule-based
        Expert System for Packaging (ESP))
  • Misunderstood Curry-Howard correspondence
  • Misunderstood relation of project to ordinal
    analysis
=== There have been many mistakes commited. --- ## SECTION Releasing the FRDCSA === --- 
Releasing FRDCSA:
 
  • Haven't released full system due to personal data
    strewn throughout the repositories
    • Repositories are dozens of GB large,
      • Must semi-automate the process
        • The scrubber/redactor/release-manager/etc systems
  • Entire system is heavily rootbound
    • Hard to factor it into orthogonal components
      • Although minor/task2 is supposed to do just that
        as needed
  • Did release a stripped down, older, terminal version of
    FRDCSA/FLP via the Panoply-Git GNU/Linux VM
=== Releasing FRDCSA. --- 
Releasing FRDCSA:
 
  • Cannot demo most functionality because of private
    data
  • Massive technical debt and disorganization since
    lacking funds to purchase hardware such as NAS
  • Have written two auto test generators, for bringing
    under test
    • Automated-Legacy-Tests
    • AOP-SWIPL
  • Need to arrange for continuity of the project
  • Logicmoo needs to be finished, hence, funded
  • Someone to manage a nonprofit
    • Since I cannot
  • Need assistance, not instruction
=== --- Just recently, some interested collaborators have helped to motivate and push me to fast-track the FRDCSA (Panoply) release. We've decided to make the outdated and redacted Panoply git codebases at https://github.com/aindilis the stable, only branch of FRDCSA. --- I will then backport as much as feasible from my private version, which I've written these last 5 years. This means that a working version of FRDCSA is now installable and LIVE (i.e. no longer a terminal branch of the project). --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/images/frdcsa-installer.png === Do note although the installer is released, it's not ready yet. It should be ready soon after this presentation. Moreover, it needs some files to be sourced properly) --- ## FIXME: include clip of the robotic book scanner file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/clips/robotic-book-scanner.mp4#t=2,17 === Some stuff that has gotten FRDCSA the most attention took a couple hours, like the project's robotic book scanner. On the otherhand, FRDCSA has taken 22 years. It will be a game changer now that the system can be installed, and people will see what FRDCSA can do. --- AI really needs a first class effort like the Manhattan project, or the Apollo project, to inaugurate AI. Perhaps I'm describing the Cyc project... I admire the technical achievements of the Cyc project. --- 
Purpose of FRDCSA:
  • FRD :=
    'universal database of all mathematical knowledge'
    • (actually a KB, not DB)
    • Maximize problem solving capabilities in a formal
      mathematical sense
    • Package https://swmath.org
    • Package all known Automated Theorem Proving systems
      and proof libraries
  • CSA generalizes FRD to also include non-mathematical
    software
=== The purpose of the FRD project is to create a universal database of all mathematical knowledge. For instance, focused crawling and packaging all the software systems referenced by swMath.org. CSA includes non-mathematical software systems. --- https://en.wikipedia.org/wiki/Computational_problem/ === We are attempting to collect algorithmic solutions to as many computational problems as we can. Thus, our software archive contains solutions to large numbers of computational problems. The problems that people routinely face are usually reducible to these computational problems. --- 
 Various kinds of computational problems:
 
  • Decision Problem
  • Search Problem
  • Counting Problem
  • Optimization Problem
  • Function Problem
=== There are more kinds of computational problems, however, than just decision problems, as mentioned at this URL. --- ## SECTION FRDCSA Theory === --- The FRDCSA theses (combined) say that it is imperative to collect and package free/libre open-source software en masse. --- https://frdcsa.org/~andrewdo/presentation-6.jpg === This is an early and partially incorrect write-up of the deductive part of the FRDCSA thesis. But the statement of the results I believe to be correct. --- FRDCSA 'Thesis' Part 1.A. (deductive):
∀ P ∈ TR (∃ Q ∈ TR (larger(Q,P) ∧ stronger(Q,P))) === Part 1 A of the FRDCSA thesis says that for any program there is always a larger and stronger program. By larger I mean greater Kolmogorov complexity, and stronger I mean proves exactly a strict superset of (true) theorems. I like to think of this result as saying "there's always a bigger fish." --- FRDCSA 'Thesis' Part 1.B. (deductive):
¬ ∀ P ∈ TR (∃ Q ∈ TR (¬ larger(Q,P) ∧ stronger(Q,P))) === Part 1 B of the FRDCSA thesis says that it is not the case that for any program there is always a stronger program that is however not larger. This essentially means programs must ultimately grow in order to solve more problems. --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/images/godelian.png === Mechanization of the corrected statement of assumptions and proofs for Parts 1.A. and 1.B. of the FRDCSA thesis has been started by sto0pkid, and is available here. --- "Can't say I've found anything that I'd call a 'defect' in your theory. Of course there were minor details in your original formulation of the logical statements that had to be sorted out but even though my Agda formalization isn't quite 100% yet I have no doubt at this point that the assertions about the necessity of aggregation are true, mathematically...." --- "...On the other hand what to actually do with that information in practice is something I'm rather less confident about." - sto0pkid --- The Part 1.A. and 1.B. follow from the fact that no program can decide the truth values of all statements in an undecidable logic. This Agda proof seems complicated to me, but the intuition is simple. --- For Part 1.A., since no program can decide all, that means that there are always more decisions that can be added to the initial program. --- For Part 1.B., the issue is that the programs must eventually become larger, since for any given program there are only a finite number of smaller programs, but there are infinite sequences of increasingly strong programs. --- FRDCSA 'Thesis' Part 2.A. (abductive): One way to increase program strength is to collect software (in lieu of a better approach). === Part 2 A of the FRDCSA thesis is abductive, and says that one way to increase program strength is to collect software (in lieu of a better approach). --- file:///var/lib/myfrdcsa/codebases/minor/flp-videos/video/resources/docs/frdcsa-thesis-part-2.txt === Here is a first try to formalize this. --- FRDCSA 'Thesis' Part 2.B. (abductive): Given the extent of free software, packaging this software is a necessary step of any practical attempt by our group to increase our solution space. --- https://frdcsa.org/~andrewdo/writings/benefits-of-making-packages.html === There are many benefits to packaging software. --- FRDCSA 'Thesis' Part 2.C. (deductive?): Collecting software is not sufficient, therefore we must write numerous systems to help us glue the functionality of the software together where necessary. --- https://frdcsa.org/~andrewdo/writings/minor-codebase-descriptions.txt === I have written over 900 codebases. However, some of them are mere stub projects, containing, for instance, notes on how to proceed. --- https://altruisticsoftware.org/frdcsa/introduction/doc/acm74.pdf === This paper by Gregory Chaitin expresses the idea particularly clearly. --- "The approach of this paper, in contrast, is to measure the power of a set of axioms, to measure the information that it contains. We shall see that there are circumstances in which one only gets out of a set of axioms..." --- "...what one puts in, and in which it is possible to reason in the following manner. If a set of theorems constitutes t bits of information, and a set of axioms contains less than t bits of information, then it is impossible to deduce these theorems from these axioms." --- Simply replace 'theorems' with 'problems', 'axioms' with 'programs', and 'deduce' with 'solve' to understand the general necessity of having more software. --- I.E.: "If a set of problems constitutes t bits of information, and a set of programs contains less than t bits of information, then it is impossible to solve these problems from these programs." --- What is the channel capacity of one human programmer typist (for meaningful information), vs millions of programmers? We must therefore rely on the work of others, in order to address the algorithmic information bottleneck. --- file:///home/andrewdo/Downloads/bear.jpeg === The project theses leave open the question of how fast program size should be increasing. I need to formalize this answer: "We can't out run this bear!! True, but I can out run you." --- 
andrewdo@ai2:~$ apt-cache search 'artificial intelligence'
cl-reversi - Reversi game for Common Lisp
python-genetic - genetic algorithms in Python
lebiniou - displays images that evolve with sound
newlisp - LISP like, general purpose scripting language
andrewdo@ai2:~$
=== There is an alarming lack of packaged Artificial Intelligence software. This example shows how the search phrase 'artificial intelligence' has but 4 matches out of 60,000 Debian packages. --- https://altruisticsoftware.org/frdcsa/git/index.html === People on the #debian channels ask - well what AI systems would you package? For starters, I have semi-automatically collected over 2000 free/libre open source A.I. systems., and maintain a prioritized list of software to make into packages. --- https://altruisticsoftware.org/frdcsa/external/index.html === --- ## FRDCSA/FLP system is a proof of concept, which nevertheless helps me ## organize in the meantime. ## --- ## [Show ArchiveBox, and NLU-MF] ## === Engineering the FRDCSA can be likened to creating a tool shop, with the intuition that it is provident to collect and prepare tools for use before problems are encountered, thus sparing the possible failure of a last-minute search for the proper tool under time pressure. --- file:///var/lib/myfrdcsa/codebases/minor/argument-system/data-git/argdown/graphs/test.argdown.png === I have begun argument mapping various debates surrounding the FRDCSA and FLP projects, in order to precisely characterize them, and hopefully generate arguments and proofs that support the goals and strategies of the project. These arguments could hopefully be used to help persuade rational people to support the project in any way they can. --- 
Acknowledgements:
Thanks to:
 
  • dmiles
  • CasAm
  • jbalint
  • koo5
  • sto0pkid
  • Justin Coslor
  • Christina Taft
  • Meredith McGhan
  • Paul Fierro
  • Dr. Ravi Sharma
--- 
# https://frdcsa.org
# https://freelifeplanner.org
# https://github.com/aindilis
# https://frdcsa.org/~andrewdo
# https://facebook.com/frdcsa
=== --- THE END