Re: KR for Human in the Loop": Two challenges related to KR..

Paola, Pete et al

Thanks for your comments and phrasing used in the modified call below..

I invite everyone to indicate your level of interest in participating in a
series of meetings :

   - Objective is to determine how to Use KR to support H*umans in the AI
   loop*
   - One task is to explain the challenges that Human in the Loop Knowledge
   Representation would  address.
   - Startpoint is the document produced during our meetings on  'Leveraging
   the StratML specification for AIKR'


Happy Thanksgiving

Carl Mattocks

It was a pleasure to clarify


On Mon, Nov 21, 2022 at 9:20 PM Paola Di Maio <paola.dimaio@gmail.com>
wrote:

> Carl, Human in the AI loop is a good idea,
>
> Human in the loop is very broad, what about being a touch more precise
>
> *Using KR to support humans in the AI loop*
> (maybe you can phrase it even better)
>
> The goals and content would have to be aligned to the title
> StratML however useful is neither specific to AI or KR
>
> i have absolutely no problem with the fact that Owen translates every
> statement
> to stratML, however this list is not about stratml at all
> apparently some members are confused by the frequency of stratml posts
> and wonder if this list is about stratml
>
> May we suggest that Owen, when kindly and cheerfully makes a stratml page
> for everything that we discuss here, refrains from making each time a
> public announcement on the list about it, and just pings the statement
> owner ?
>
> Owen of course you are very welcome to continue to contribute to all
> discussions, but maybe we do not need to be informed everytime you make a
> stratml entry?
> what do you think? :-)
> I think there may be scope for using stratML to make explicit statements
> about AI, each AI could have a stratML like schema to declare what it does
> and how it does it
> however if I remember correctly you said you have no plan to modify
> stratML at the moment
>
>
> On Tue, Nov 22, 2022 at 4:28 AM Peter Rivett <
> pete.rivett@federatedknowledge.com> wrote:
>
>> Hi Carl,
>> I don't know if it's a copy-and-paste error but I don't see how the title
>> "KR for Human in the Loop" matches the objective which is about the
>> somewhat legacy XML language StratML; which AFAIK is for strategic
>> performance planning as opposed to AI, human involvement in AI, or
>> knowledge representation except for the very narrow domain of knowledge of
>> strategic plans.
>>
>> Apologies for missing background from previous pre-COVID discussions, but
>> I'm sure I won't be the only one: are there any archives or outputs?
>> Maybe an explanation of the specific problem space related to Human in
>> the Loop Knowledge Representation would help: for example the competency
>> questions it's hoped to address.
>>
>> Regards
>> Pete
>>
>> Pete Rivett (pete.rivett@federatedknowledge.com)
>> Federated Knowledge, LLC (LEI 98450013F6D4AFE18E67)
>> Schedule a meeting at https://calendly.com/rivettp
>>
>>
>> ------------------------------
>> *From:* carl mattocks <carlmattocks@gmail.com>
>> *Sent:* Monday, November 21, 2022 10:29 AM
>> *To:* Dave Raggett <dsr@w3.org>; W3C AIKR CG <public-aikr@w3.org>
>> *Cc:* Stanislav Srednyak, Ph.D. <stanislav.srednyak@duke.edu>
>> *Subject:* KR for Human in the Loop": Two challenges related to KR..
>>
>> KR Folk
>>
>> To Give a measure of Thanks at this time of Thanks Giving .. I invite
>> members to show their level of interest in participating in a regular
>> conference call to discuss "KR for Human in the Loop"
>>
>> Objective is to continue defining  how "StratML" helps explain AI KR.
>> Specifically, before Covid, we had mapped out how "Human in Loop" was a
>> significant factor in shaping use of AI KR .. But we had no "language" for
>> that interaction.
>>
>> Cheers
>>
>> Carl Mattocks
>> CarlMattocks@WellnessIntelligence.Institute
>> It was a pleasure to clarify
>>
>>
>> On Mon, Nov 21, 2022 at 6:05 AM Dave Raggett <dsr@w3.org> wrote:
>>
>> If you want an natural language notation for math, you might be
>> interested in EasyMath from work in the late nineties:
>>
>> EzMath provides an easy to learn notation for embedding mathematical
>> expressions in Web pages. The notation is inspired by how expressions are
>> spoken aloud together with a few abbreviations for conciseness (e.g. x^y
>> denotes x raised to the power y).
>>
>>
>> See:
>>
>> https://www.w3.org/People/Raggett/EzMath/
>> https://www.w3.org/People/Raggett/EzMath/EzMathPaper.html
>>
>> Sadly, the browser plugin is now defunct as it relies on an interface
>> long abandoned by modern browsers.  It wouldn’t be that hard (one week's
>> work) to reimplement it as a JavaScript library using the HTML CANVAS
>> element as its target.
>>
>> However, that is a million miles from work on AI agents like Minerva.
>>
>> Minerva is a sophisticated deep learning based system. It starts from
>> general purpose large language model (PaLM) and refines it with training
>> against a mathematical dataset, producing impressive results.
>>
>> https://arxiv.org/pdf/2206.14858.pdf
>>
>> However, the approach described in the paper (linked above) is limited to
>> agents with a single purpose. For agents designed for general purposes, we
>> need a more flexible approach. That is why I am proposing work on direct
>> manipulation of latent semantics, along with mimicking the way that the
>> brain separates different kinds of knowledge across different parts of the
>> cortex. The idea is to combine intuitive (System 1) thinking with
>> deliberative, analytic thinking (System 2).  Minerva only supports the
>> former.
>>
>> On 21 Nov 2022, at 10:00, Paola Di Maio <paola.dimaio@gmail.com> wrote:
>>
>> You and I are on different planets, and speak different languages :-)
>>
>>
>> So it seems. :-)
>>
>> Dave Raggett <dsr@w3.org>
>>
>>
>>
>>