Re: group's quick updates, and have a nice summer all

> On 15 Aug 2022, at 16:45, ProjectParadigm-ICT-Program <metadataportals@yahoo.com> wrote:
> 
> Dear Paola,
> 
> Is there a consensus on what constitutes neurosymbolic knowledge representation? Can you list the documents for review in order for us to propose and upgrade to WG?

Indeed, as proposing a WG seems rather premature.  Drafting a WG charter is something to do when there is a mature body of work and clear industry support for the proposed direction.

Did you see: Neuro-Symbolic Approaches for Knowledge Representation in Expert Systems, 2004, Ioannis Hatzilygeroudis and Jim Prentzas? I reckon that that work is rather dated, but it was placed early in the search results for neurosymbolic knowledge representation! 

> I think the idea is fine, but we need to create a proposal that covers more than just AI, by looking how AI is used in various fields to deal with formalizing and representing knowledge.
> 
> High energy physics, conformal field theory, quantum physics, theoretical and elementary particle physics, biochemistry, computational linguistics, neuroscience, mathematics and machine learning.

What about cognitive science, sociology and psychology? I would expect to include support for human cognition and natural language semantics at the very least. This is something I am progressing with my work on plausible reasoning.  It builds upon symbols together with qualitative parameters and emulating human memory, along with unconscious and conscious thought.

> All of these use common concepts and formal systems and knowledge representation systems. If we can find a set of common concepts and knowledge representation systems , it will be much easier to come up with a viable proposal.

I wouldn’t insist on formal systems as we are still at an early stage of understanding human cognition and premature formalisation is likely to act as a brake on open thinking. If you have invested in an expensive hammer, then everything looks rather like a nail.

> I find the term neurosymbolic a bit limitative. What we are looking for is interchangeable (homomorphic) KR systems that can be mapped to cognitive architecture or neuroscience based modeling.

Quite so, apart from your use of “homomorphic”.  I very much support the idea of distinguishing between functional high level representations and the multitude of potential lower level representations, including, cognitive databases, noisy vector spaces, pulsed neural networks and so forth, each of which will have pros and cons.

p.s. I have recently expanded my demo to include analogical reasoning in preparation for a plenary talk I am giving next month at the workshop on analogies from theory to applications, see: https://iccbr-ata2022.loria.fr <https://iccbr-ata2022.loria.fr/>. My web-based demo can be found at: https://www.w3.org/Data/demos/chunks/reasoning/ <https://www.w3.org/Data/demos/chunks/reasoning/>. I am now busily working on implementing the algorithms that deal with computing likelihood given a suite of qualitative parameters, inspired by pioneering work by Alan Collins and his colleagues in the 1980’s.  

> 
> I will send a couple of links to interesting takes on knowledge and knowledge representation.
> 
> 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 Monday, August 15, 2022 at 02:17:58 AM AST, Paola Di Maio <paola.dimaio@gmail.com> wrote:
> 
> 
> 
> Hope everyone is having a good summer and getting lazy time
> and welcome to new members
> Please share your version of the classic summertime tune, if you have any <https://www.youtube.com/watch?v=PVUOtLLGv2w&ab_channel=DennisTschirner%7CVocalist> 
> 
> I was hoping to be totally away from the computer, but got busy catching up deadlines, I am however managing to spend time on the beach almost every day, which means I am at least half on summer holiday. I hope this much for you all, at a minimum.
> 
>  As you may have noticed, I am having trouble with my eyesight. Spell checker does not catch everything. Apologies for the typos. I am not just being sloppy.
> 
> AI KR standards are badly needed, and W3C is on the case
> 
> We plan to advance this CG to WG, so that we can draft a standard
> for neurosymbolic knowledge representation, NSKR, at some point soon.
> 
> what is NSKR you may ask-
> 
> I had to carry out a lot of research and argue fairly hard for NS KR because obviously some people say that NSI and KR are two different things, also because I had not yet fully worked it out.
> 
> I have now completed the background research to  propose a standard for neurosymbolic integration using model cards, and illustrated the case with citable references (in the hope it will be cited, not plagiarized)
> 
> I needed to have some well thought out arguments and references before I could  invite members of this AI KR CG to think about upgrading to WG and start working on a standard 
> 
> I ll soon be sharing the main points of the plan, and provided enough members are ready to receive, well need 5 members to promote the CG to WG and then we can get working
> 
> Finally, the first report was uploaded today - same old report presented at TPAC 2021, which basically summarises the main arguments that motivated this CG and point to future developments - just updated it it a bit since drafting
> https://www.w3.org/community/aikr/wiki/File:AI_KR_FIRST_REPORT_PUBLISHED_VERSION.pdf <https://www.w3.org/community/aikr/wiki/File:AI_KR_FIRST_REPORT_PUBLISHED_VERSION.pdf>
> Apologies for taking so long. Future reports may be published more timely if someone volunteers to proofread.
> 
> Richard Lea was nominated cochair last spring and has agreed to help set up a github repo for AI KR, he is working on it 
> He has also agreed and to work towards forming a WG so that we can move on towards devising a new standard
> 
> Lastly, I have never really considered using the Internal mailing list
> internal-aikr@w3.org <mailto:internal-aikr@w3.org>  until Daniela Cialfi, a new member, posted there recently
> (Monday, 25 July) Thank you Daniela. I guess there are too many trolls on the open web. <>
> The internal-AIKR mailing list is only accessible to authenticated users, rather than being publicly accessible. It could be a good idea to post to the internal list only anything that group members are not yet ready to share without authentication *referring to plagiarism concerns discussed earlier
> 
> Continue to have a good summer, to resist  radical connectionism in machine learning and more about the proposed standard soon
> 
> Paola Di Maio
> Chair W3C AI KR CG

Dave Raggett <dsr@w3.org>