- From: ProjectParadigm-ICT-Program <metadataportals@yahoo.com>
- Date: Sun, 14 Jun 2020 22:25:59 +0000 (UTC)
- To: W3C AIKR CG <public-aikr@w3.org>, carl mattocks <carlmattocks@gmail.com>
- Message-ID: <1821802580.513904.1592173559389@mail.yahoo.com>
We seem to be forgetting something very important. Knowledge itself is context sensitive, perception sensitive and paradigm sensitive. I would be very careful in what we define to be AI KR objects, because we again return to objects as a focal point.
By doing so we limit the knowledge domains covered, and because AI has as a distinguishing feature the capacity to learn, we must also look at the different forms of knowledge, corresponding philosophical, perceptual frameworks and how conceptualization is captured in natural language and or formalized (logical) structures, which can be using symbols or visual tools, such as diagrams, flowcharts etc.
And because we are in the business of defining ethical use of AI we definitely need to look at all the forms of knowledge.
Milton Ponson
GSM: +297 747 8280
PO Box 1154, Oranjestad
Aruba, Dutch Caribbean
Project Paradigm: Bringing the ICT tools for sustainable development to all stakeholders worldwide through collaborative research on applied mathematics, advanced modeling, software and standards development
On Sunday, June 14, 2020, 1:36:00 PM ADT, carl mattocks <carlmattocks@gmail.com> wrote:
To restate in simpler language - AI KR Object may be :
- an Algorithm (example - enable an entity to determine consequences)
- an Ontology (which has a set of ontological commitments)
- an Intelligent Reasoning (fragmentary) Theory, such as, deduction, induction, abduction, by analogy, probabilistic, case-based
- the Reasoning Mechanism (computational environment), such as, natural language processer, rules engine, machine learning
- a Vocabulary (medium of human expression) e.g. StratML, RuleML
It was a pleasure toclarify
On Sat, Jun 13, 2020 at 10:47 AM carl mattocks <carlmattocks@gmail.com> wrote:
Regarding AIKR object -
Object is a resource as per the https://www.w3.org/RDF/ Resource Description Framework (RDF) which provides a basic capability to define classes, subclasses, and properties of objects..Other technologies, like OWL or SKOS, build on RDF and provide language for defining structured, Web-based ontologies
Acknowledging that definition of AI and KR continue to evolve .. and some challenges persist .. as described in the 1993 paper (attached) titled What Is a KnowledgeRepresentation? which argued that a knowledge representationplays five distinct roles, each important to thenature of representation and its basic tasks ..AND .. suggests that combiningrepresentations is a task that should be drivenby insights about how to combine their theories of intelligent reasoning, not their implementation mechanismshttps://en.wikipedia.org/wiki/Knowledge_representation_and_reasoning#cite_note-12
First, a knowledge representation is mostfundamentally a surrogate, a substitute for thething itself, that is used to enable an entity todetermine consequences by thinking ratherthan acting, that is, by reasoning about theworld rather than taking action in it. Second, it is a set of ontological commitments, that is, an answer to the question, Inwhat terms should I think about the world? Third, it is a fragmentary theory of intelligent reasoning expressed in terms of threecomponents: (1) the representation’s fundamental conception of intelligent reasoning,(2) the set of inferences that the representation sanctions, and (3) the set of inferencesthat it recommends. Fourth, it is a medium for pragmaticallyefficient computation, that is, the computational environment in which thinking isaccomplished. One contribution to this pragmatic efficiency is supplied by the guidancethat a representation provides for organizinginformation to facilitate making the recommended inferences. Fifth, it is a medium of human expression,that is, a language in which we say thingsabout the world.
enjoy
Carl
It was a pleasure toclarify
On Fri, Jun 12, 2020 at 8:25 AM carl mattocks <carlmattocks@gmail.com> wrote:
We will be addressing Evaluation in the discussions around use of (StratMl part 2) KPIs which are essential for measuring success of a strategy.
Agreed - KR activities can start only after the goal and scope of the AI system have been defined
- Document the vision, values, goals, objectives for one or more AIKR objects
- Employ ontological statements when explaining AIKR object audit data, veracity facts and (human, social and technology) risk mitigation
It was a pleasure toclarify
On Thu, Jun 11, 2020 at 8:46 PM Paola Di Maio <paola.dimaio@gmail.com> wrote:
However, we can develop KR say ontologies, which are implementation independent, Perhaps look at the role of the ontologist, to get a sense of the role for the AI KR strategist?
I stress that I dont think of this as a strategy but adopting this construct as it is part of the stratml vocab and can be useful way of of looking at itWhen it comes to ML, as per your table (ML Evaluation) it is definitely not the role of the AI KR strategist to evaluate machine learning imhoTrack (monitor) and evaluate are part of the same task imho
It could be the role of the AI strategist to decide whether ML is needed in the AI, and to evaluate the alternativesThe role of the AI KR strategies would be to devise and implement the KR according to the type of AI design/methods being used by the system.
The evaluation of a ML techniques irequires different approaches from the evaluation of NL systems, for example
On Fri, Jun 12, 2020 at 8:23 AM Paola Di Maio <paola.dimaio@gmail.com> wrote:
Thank you for clarifying. The table is neat and a good starting point for elaboration. It is a bit scattered but that can be sorted.
I dont see KR though
1. Can you please point out which of these items fall under KR? (say, ontology?)
my experience, the AI KR development work starts after the case for the AI system has already been made, , ie, when the need for the AI system has already been identified, ( the first two points). The system design is in place schedules and resources to be in place,KR activities are not AI general system development activities. What I mean is that KR activities can start only after the goal and scope of the AI system have been defined. (not sure whose role is to define them, perhaps the system developer or the AI strategist)
On Thu, Jun 11, 2020 at 10:44 PM carl mattocks <carlmattocks@gmail.com> wrote:
Please peruse this GOAL Dependency approach to sequencing
|
Goal Activities
|
Title AI Strategist Goal
|
Goal Dependency
|
AIKR Strategist Protocol
|
|
Give an understanding of the possible applications of AI to conversations/decisions about business strategy
|
Business Strategy
|
Document, Requirements, Quality, Robust, Ontological Statements, Ethics, Lawful, Machine Learning Evaluation, Track
|
yes
|
|
Identify which areas of the requirements warrant AI solutions versus which can be achieved with other types of solutions
|
Requirements
|
Business Strategy, Document, Quality, Robust, Ontological Statements, Ethics, Lawful, Machine Learning Evaluation, Track
|
yes
|
|
Document the vision, values, goals, objectives for one or more AIKR objects
|
Document
|
Business Strategy, Requirements, Quality, Robust, Ontological Statements, Ethics, Lawful, Machine Learning Evaluation, Track
|
yes
|
|
Define the limits of quality. If a product has limits of quality/action, then these should be stated
|
Quality
|
Business Strategy, Document, Requirements, Robust, Ontological Statements, Ethics, Lawful, Machine Learning Evaluation, Track
|
yes
|
|
Evaluate machine learning models
|
Machine Learning Evaluation
|
Business Strategy, Document, Requirements, Quality, Robust, Ontological Statements, Ethics, Lawful, Track
|
yes
|
|
Track AIKR object performance outcome via KPI (Key Performance Indicator) based on supervised learning models measurements
|
Track
|
Document, Requirements, Quality, Robust, Ontological Statements, Machine Learning Evaluation
|
yes
|
|
Ensure AI Systems are designed to handle uncertainty and tolerate perturbation from a likely threat perspective, such as, design considerations incorporate human, social and technology risk factors
|
Robust
|
Document, Requirements, Quality, Ontological Statements, Ethics, Lawful, Machine Learning Evaluation, Track
|
yes
|
|
Employ ontological statements when explaining AIKR object audit data, veracity facts and (human, social and technology) risk mitigation
|
Ontological Statements
|
Business Strategy, Document, Requirements, Quality, Robust, Ethics, Lawful, Machine Learning Evaluation, Track
|
yes
|
|
Ensure AI Systems adhere to principles of ethics
|
Ethics
|
Document, Requirements, Quality, Robust, Ontological Statements, Lawful, Machine Learning Evaluation, Track
|
yes
|
|
Ensure AI Systems comply with all applicable laws and regulations, such as, provision audit data defined by a governance operating model
|
Lawful
|
Document, Requirements, Quality, Robust, Ontological Statements, Ethics, Machine Learning Evaluation, Track
|
yes
|
Carl Mattocks
It was a pleasure toclarify
On Wed, Jun 10, 2020 at 8:46 AM carl mattocks <carlmattocks@gmail.com> wrote:
Please find attached (summary extract below) Strategic Plan - The role of the AI KR Strategist (A strategist is a person with responsibility for the formulation and implementation of a strategy).
To help validate the list of goals we invite proposals for a sequencing (of the goals) that could be the foundation of an AI Strategy Protocol.
Please provide your proposals as a reply to this post.. intent is to apply them to the StratMl version at 9:00 am (NY time) June 23 meeting.
thanks
Carl MattocksAIKRCG co-Chair
Strategic Plan
The role of the AI KR Strategist
For: Artificial Intelligence Knowledge Representation Community Group(AIKR CG)
SubmittedBy: Carl Mattocks
y CarlMattocks@WellnessIntelligence.Institute
Overview
This plan defines the role of the AI KRStrategist.
Contents
No table of contents entries found.
1. Articulation
1.1. Vision
For all AI systems to have clearly andtransparently documented goals and performance data showing that they are beingachieved.
1.2. Mission
The mission of an AI Strategist is todefine the purpose and goals of AI systems, as well as the KPIs by which we candetermine if the system is meeting its goals.
1.3. Scorecard
|
|
Goals
|
Objectives
|
Performance Indicators
|
Commentary
|
|
Goals with no perspective
|
Quality (see pp1)
|
Adherence to Environmental Impacts (see pp1)
|
|
|
|
|
|
Efficiency (see pp1)
|
|
|
|
|
|
Ethics (see pp1)
|
Accountability (see pp1)
|
|
|
|
|
|
Autonomy (see pp1)
|
|
|
|
|
|
Confidentiality (see pp1)
|
|
|
|
|
|
Veracity (see pp1)
|
|
|
|
|
|
Machine Learning Evaluation (see pp1)
|
Track (see pp1)
|
Precision Recall
|
|
|
|
|
Accuracy
|
|
|
|
|
Confusion Matrix
|
|
|
|
|
Per-class accuracy
|
|
|
|
|
Log-Loss
|
|
|
|
|
AUC-ROC Curve
|
|
|
|
|
F-measure
|
|
|
|
|
NDCG
|
|
|
|
|
Regression Analysis
|
|
|
|
|
Quantiles of Errors
|
|
|
|
|
"Almost correct" predictions
|
|
|
|
|
Trustworthy (see pp1)
|
|
|
|
|
|
Lawful (see pp1)
|
|
|
|
|
|
|
Ontological Statements (see pp1)
|
|
|
|
|
|
|
Track (see pp1)
|
|
|
|
|
|
|
Business Strategy (see pp1)
|
|
|
|
|
|
|
Requirements (see pp1)
|
|
|
|
|
|
|
Document (see pp1)
|
|
|
|
|
|
|
Robust (see pp1)
|
|
|
|
|
|
1.4. Goals....
It was a pleasure toclarify
On Tue, Jun 9, 2020 at 11:56 PM Paola Di Maio <paola.dimaio@gmail.com> wrote:
oops-Thanks Carl Owen and Chrisfor capturing and sharing.I repeat here my comment shared with participants after the session - also pasted below
if the plan we worked on yesterday is to describe the role of an AI Strategistthen we need to make explicit what the goal of the role and associated activities is/are
AI systems /software development can be very sophisticated and well understood practicein systems/software engineering . It is not a strategy role to do requirements analysis and writetechnical systems specifications, especially not technically advanced ones So we should review and devise an AI strategist/strategy goalsaccordingly, taking into account the technical system specifications.or viceversa?A strategist could identify the need for a technical system based on a business problem/need, but then pass the developmentto a specialist (a systems developer/engineer) who has the skills to do soSo upon further reflection, I dont think requirements belong to an ai strategy. unless better specified
pdmpaola wrote:Thanks for starting this planAs I read what had been drafted I commented with my software/system developer hat onThe system lifecycle is considered a process, rather than a strategy, and consists of identifying requirements, specifications and documentation and allIts well understood and documented and there are several models, but the lifecycle itselfis fairly standard, it applies to AI systems as to any systemSo before I have a go at reorganising and perhaps streamlining a bit those goalsI think we should define better what is the role of the AI strategistdoes it overlap the role of a ai system analyst and designer (which is what I have in mind)ie identify what needs to be doneor is it more a role to decide HOW these goals need to be achieved interms of resources, schedules, policies, coordination with management etcAlso, as I mentioned, this may depend on what other roles the AI strategist works with/depends on- p
On Wed, Jun 10, 2020 at 11:45 AM Paola Di Maio <paola.dimaio@gmail.com> wrote:
On Wed, Jun 10, 2020 at 12:38 AM carl mattocks <carlmattocks@gmail.com> wrote:
Role of Data Scientist has a process that is described in the links below :
Cross-industry standard process for data mining, known as CRISP-DM,[1] is an open standard process model that describes common approaches used by data mining experts. It is the most widely-used analytics model
https://en.wikipedia.org/wiki/Cross-industry_standard_process_for_data_mining
these lecture notes provide good background notes
https://wiwi.hs-duesseldorf.de/personen/thomas.zeutschler/Documents/HSD_W_ITAiBA_Zeutschler_SS2016_Lecture2_CRSIP_DM.pdf
this is an IBM presentation on the The Data Science Processhttps://www-01.ibm.com/events/wwe/grp/grp304.nsf/vLookupPDFs/Polong%20Lin%20Presentation/$file/Polong%20Lin%20Presentation.pdf
enjoyCarl
It was a pleasure toclarify
On Tue, Jun 9, 2020 at 12:13 PM Owen Ambur <Owen.Ambur@verizon.net> wrote:
Carl, thanks for leading another productive editing session on the AIKR CG televideo conference today. I found myself agreeing with much of what Paola was saying, and I continue to be impressed by Chris' capability to comprehend and capture the essence of our dialog in his StratNavApp. For the benefit of those who were unable to join us, I am providing a link here so that they can view our draft plan as it currently exists.
I'm sorry that Justin was not able to participate to ensure that we are capturing the knowledge he has gained in considering the roles of AI strategists. However, it would be great if he and others could contribute their comments and edits on our draft in Chris' app and perhaps participate in our text televideo conference on June 23.
While the StratML collection does not yet contain a model performance plan for Data Scientists, it does include one for Chief Data Officers (CDOs), at https://stratml.us/drybridge/index.htm#PP4CDO
Unless someone has a better source, I will soon convert the contents of this one to StratML format: https://www.sas.com/en_us/insights/analytics/what-is-a-data-scientist.html
BTW, Carl, your reference to OASIS's Business Centric Methodology (BCM) prompted me to convert its content to StratML format at https://stratml.us/drybridge/index.htm#BCM
These elements of the BCM are closely related to the purposes of the StratML standard:
Objective 1.1: Goals - Determine the business goals.
Objective 1.3: Participants - Identify the project participants.
Objective 1.4: COIs - Identify the Community of Interest. Goal 2: Goals - Understand the business goals.
Objective 3.3: Standards Bodies - Map interoperability requirements to standard bodies. Objective 3.4: Formats - Map interoperability requirements to internal legacy system formats.
Objective 3.5: Templates - Capture information in a way that can be reused over time and among participating organizations. Objective 3.5.1: Strategies - Implement strategies for interoperability. Objective 3.5.5: Standards - Apply standards for interoperability and contracts. Objective 3.5.7: Performance - Monitor and manage performance.
Objective 4.1: Standards - Use existing standards.
It would be great if BCM practitioners were to use the StratML standard (ISO 17469-1). Anyone who may wish to apply the BCM template as a performance plan for their organization in StratML Part 2 format could do so by clicking on the link provided here.
Owen
-------- Forwarded Message --------
| Subject: | Re: ai strategist role |
| Date: | Tue, 9 Jun 2020 11:22:53 -0400 |
| From: | carl mattocks <carlmattocks@gmail.com> |
| To: | Paola Di Maio <paoladimaio10@googlemail.com> |
| CC: | Owen Ambur <Owen.Ambur@verizon.net>, Chris Fox <chris@chriscfox.com>, Paul Alagna <pjalagna@gmail.com> |
Data Scientist is a complimentary role.. which is supported by an open source method that has strong emphasis on machine learning and is Plan Do Check Act oriented.
Carl
On Tue, Jun 9, 2020, 11:12 AM Paola Di Maio <paola.dimaio@gmail.com> wrote:
Thanks for starting this plan As I read what had been drafted I commented with my software/system developer hat on The system lifecycle is considered a process, rather than a strategy, and consists of identifying requirements, specifications and documentation and all Its well understood and documented and there are several models, but the lifecycle itself is fairly standard, it applies to AI systems as to any system So before I have a go at reorganising and perhaps streamlining a bit those goals I think we should define better what is the role of the AI strategist does it overlap the role of a ai system analyst and designer (which is what I have in mind) ie identify what needs to be done or is it more a role to decide HOW these goals need to be achieved in terms of resources, schedules, policies, coordination with management etc Also, as I mentioned, this may depend on what other roles the AI strategist works with/depends on- p
Received on Sunday, 14 June 2020 22:26:15 UTC