from reliability engineering to trustworhy AI, transposing concepts and terms

I have finally boiled down a rationale and long list of terms that in my
understanding is representative of the KR domain
 for the purpose having a first stab at the AI KR vocabulary,  Keeping in
mind that my concern is reliability/trustworthyness

 I ll soon share it for evaluation and discussion

I want to apologise to everyone who tried to figure out what I have been
working on and how from my messy messages
I assure you I can explain -)

 I had some off list exchanges trying to figure out what would be the best
knowledge sources to start extracting terms from
It has been an epistemological challenge - because KR is so vast and has so
many facets

I d like to share a paper that explains how 'reliability engineering'
concerns can be transposed to trustworthy AI
*the correctness of the AI system is a priority for KR
So in addition to a bunch of terms /concepts from related domains
*Description Logic for example  I am including in our AI KR
terms from reliability engineering,

Afef Awadid, Kahina Amokrane-Ferka, Henri Sohier, Juliette Mattioli, Faouzi
Adjed, et al..
 *AI Systems Trustworthiness Assessment: *State of the Art. Workshop on
Model-based System Engineering
and AI, 12th International Conference on Model-Based Software and Systems
Engineering (Model-
sward), Feb 2024, Rome, Italy. �hal-04400795�

https://hal.science/hal-04400795/document



---------- Forwarded message ---------
From: Peter Rivett <pete.rivett@federatedknowledge.com>
Date: Sat, Dec 7, 2024 at 5:28 AM

To: Chris Harding <chris@lacibus.net>, Paola Di Maio <
paoladimaio10@gmail.com>
Cc: John F. Sowa <sowa@bestweb.net>


Good progress, though I'd add some concerns...

   1. I'm not sure our objective is useful "The minimal set of concepts
   necessary for an automated system (say any
   >     AI) to function correctly (and not spuriously or incorrectly)
   which are
   >     not understood/captured/defined by existing AI standards"
   characterises
   >     what we are looking for.


   1. In general an AI needs to understand concepts related to its problem
   area not AI generally. The vocabulary we're building I'd say is general,
   and I'd say its purpose is for "discussing, comparing, evaluating and
   characterizing AI systems themselves" which is quite different.


   1. B ) We need to somehow decide which reference work should be
   considered authoritative for which type of term. For example I wouldn't say
   John Sowa is very definitive for "object oriented system" and that's
   reflected in the rather limited definition (IMO, speaking as someone well
   versed in OO)

C) we need to beware merging multiple definitions into a single one - since
we'll end up with something that is not quotable or attributable. Worse, it
could munge two quite different meanings e.g. "bank: a financial
institution located on the side of a river or stream". Better, I think, to
keep the definitions separate, maybe with a score if that can be automated,
for subsequent selection by human expertise

Pete

Pete Rivett (pete.rivett@federatedknowledge.com)
Federated Knowledge, LLC (LEI 98450013F6D4AFE18E67)
tel: +1-701-566-9534
Schedule a meeting at https://calendly.com/rivettp

------------------------------
*From:* Chris Harding <chris@lacibus.net>
*Sent:* Friday, December 6, 2024 10:09 AM
*To:* Paola Di Maio <paoladimaio10@gmail.com>
*Cc:* Peter Rivett <pete.rivett@federatedknowledge.com>; John F. Sowa <
sowa@bestweb.net>
*Subject:* Re: John Sowa s Book. attached

Thanks, Paola, see in line below.

O