General Update, Browser Extension & HDF-5 - experimental works.

Hi All,

There's been a few things I've been working on, which are loosely coupled
with the W3C Human Centric AI works - whilst now therefore openly
discussed, hopeful for positive feedback.

Some of these notes refer to works that I wasn't sure about whether to note
in the w3c cg context.  My / The deliberation being, about seeking to focus
efforts more specifically on seeking to support the discussion and
consequential works about defining what "Human Centric AI" specifically,
is...  Focused on the safety protocols; that should support freedom of
thought, portability, safety, etc.  whilst my methodologies that seek to
pursue the honourable approach, as best I can...  which is complex, as I
think some works - could be privately supported, thereby - through
availability of more resources, improving my ability to support these
efforts; yet, also, i'm dedicated to seeking to ensure support for human
rights. As the international stuff has progressed, I've become increasingly
concerned about whether deliberations, at that level, may be more narrow
than could otherwise be the case, if alternatives now embryonic were more
demonstrable.

I'm not sure who else involved is entirely unpaid, and doing work in the
interests of humanity, morally, with limited means - but it is hard.
Whilst this problem does need to be addressed, the importance of ensuring
future ICT systems have the means available to them to support 'selfhood';
or more broadly - human centric ai - is an important underpinnings, at
least as an option that's able to be made available - so it seems, the work
to get that done, is the first priority, whilst being mindful of the
broader range of problems that designs should also seek to ensure are
addressible in future, with the intent to ensure they're actually able to
be addressed...  as an option, able to be supported by market-based
providers who choose to, noting, i think there's enormous socioeconomic
merit discoverable should that capacity be brought about...  but, the devil
is in the detail, imo.

These things are seemingly inter-twined, so, whilst I'm not seeking to
conflate issues, my mind is presently struggling to decouple aspects -
whilst being cognizant of timelines broadly elsewhere, world-wide
(particularly Re: UN GDC[1] & related works)


*Community Development*Personally, there are many communities I have
historically set-up and/or participate(d) in.  The ability to support
knowledge dissemination is problematic, as I neither want to 'spam' people,
nor be neglectful, whilst only having so many hours and resources
available.


*Human Centric AI*The historical works on 'human centric' methodologies via
Credentials CG, always related to AI[2]. My works, as an individual, whilst
difficult - provided me the means to study and work on use-cases relating
to human beings - stuff that historically, it seemed neither private nor
public entities wanted to fund work to address, as such, they're often
still not addressed.  Nonetheless, there are various design implications
that I consider important, and certainly as international works start to
consider human rights, there are demands upon how to communicate these
considerations in different ways, for different audiences, that require
tooling...  imo.

I believe, between some, there is a level of implicit consensus...  but
rarely, explicit, and certainly not explicitly universal.  it seems the
need to support multi-modal structures is important, and therefore
portability between these different types of systems, like osx vs. windows
vs. linux, only different; therein also, communications interoperability
standards; royalty free, whilst ideally also able to address 'digital
slavery' amongst other human rights related threats, but again, perhaps not
universally.

just ensuring that the availability of systems that can do that, can and do
exist, as should be allowed.

It appears as though some good conversation has been had about Solid being
an interoperability foundation at https://2023.mydata.org/ - and I look
forward to hearing more about that; and will report back, when / if that
information becomes available.  perhaps indeed, it already is, and i just
haven't seen it yet.  idk.

- Web 3 vs. web 3.0 - the historical meaning.
until recently, as in, the last couple of years - the historical concept of
'web 3.0' was essentially about the 'social web', or 'socially aware cloud
storage'[3], and similar, built upon the 'web of data' (semweb), forming an
informatics layer that amongst other things - could support - human centric
- informatics.  Perhaps, there is now a need for a new term, that can
better support & dignify the intended meaning of the pre web3 works on web
3.0...  + updates, extending from that mind-model related framework...

As may be clear to some, this involved the use of 'ai' technologies, namely
semantic web, but more broadly, denote a historic, yet not well known role
of w3c efforts, to define open-standards for AI as part of the broader
internet governance frameworks[4][5].  It appears as though internet
governance is a topic, being reviewed via the aforementioned UN works.
There's a few considerations I've had through my participation in these
events (inc. also WSIS2023);
- the level of understanding or comprehension of technology, at the 'design
capacities' level, appears low.  It appears the comprehension of the
adjacent possible, is based upon what exists in the market as a product
useful for consumers now.

Three are various other 'stakeholder analysis' related considerations /
deliberations, but none, seemingly as important as seeking to ensure that
their efforts are well-informed at the level to which they are reasonably
able to be expected to understand.

- Support for Language of Prayer, et.al.

There are a variety of areas of importance for members of our human family
at large that have not been supported well enough historically.  Some are
deeply personal, private, innate. relating to spirit, soul, selfhood...
 this situation is deeply unfortunate and must be addressed.  not doing so,
gives rise to 'splinter-net' as it's called, imo...  and/or other unwanted
outcomes, due to a lack of thoughtfulness; At least, that's my concern.

- Web of meaning - of words & category theory
Part of the implicit notions described lightly below re: sense, is
associative to the objective desire to define graphs for words that denote
the intended meaning of that word.  The methodology that I'm exploring to
resolve this issue has other use-cases.

- Health industry targets & payments

It appears to me that there is a health-industry use-case focus that
in-turn also delivers the infrastructure needed for payments.  The need for
health-care systems is legal infrastructure as health records are
in-effect, legal records, amongst the highest-stakes use-cases known to me
- something I've historically sought to avoid, but have now, in light of
what has happened over the last few years, my position has changed.
Health-Care infrastructure, has an array of different needs, but certainly,
gets to the heart of purpose for good - reality check tech, including but
not limited to - comprehensive temporal inferencing.

My observations have illustrated to me, that presently systems do not take
into account nor provide support for providing practitioners information
about legal events (ie: court-documents, etc.), nor health tracking
interfaces (ie: smart-watches / health monitors); and there are many other
sensitive use-cases, impactful to life & therefore impactful to health and
health-assessment processes (ie: OSCE[6]), that are presently not clearly
supported; Furthermore, due to the combination of,
-  the availability of these systems, the lack of comprehensive technical
(ICT/AI) capabilities of clinicians (and practices),
- the re-use of information for purposes not part of the original intended
purpose of those records, various behavioural and conflict (ie: mistake)
resolution practices vs. insurance related expectations; and
- the difference between 'claims' and 'facts', as may be found overtime-
noting, for instance, court-outcomes do not happen on the day a claim is
made; and overall,
- the engendered desire to rely upon 3rd party 'verifiable claims' rather
than patient-practitioner relationships, engenders many concerns.

The greatest concern, personally, is that the outcomes may not be good for
patients or clinicians, of which, whilst far more than real-world AI
experts, is still only a few per 1000[7] or less, depending on where.  With
the rapid advancements in AI & related fields (ie: QM, etc.) perhaps also,
there's some insights that may be gained about how to improve the number of
AI experts per 1000 for Human Centric AI related - local support
infrastructure...  nonetheless, it appears there's alot that needs to be
done.


*Sensitive use-cases*I've started looking into how 'sensitive use-cases'
may be better addressed, using some sort of system that can process - with
support for privacy - the terrible issues, "user-stories" relating to
use-cases, that are not believed to be reasonably laundered on a public
mailing list - in-order to ensure the underlying issues are addressed as
use-cases of importance, for works to define 'fit for purpose' technology.

The belief is some sort of custom, 'generative ai' system that can process
the use-cases from private truth-telling stories, into publicly addressable
use-cases.  I am not aware of how these sorts of issues have been properly
addressed earlier, and perhaps there's been consequences to the absence of
a solution to do so.  In-order to address the UN SDGs, fit for purpose
tech, should address the problems.

One of the perceived problems about how to do this is that the tooling to
do so would have other applications, such as whistleblowers and various
other related applications / implications.  Whilst I think forming a means
to do so is imperative to ensuring these works are done well, I'm still
trying to figure out the 'how' and indeed also, where the resources to do
so properly would come from.

Part of the 'human centric' paradigm, squarely, to me, relates to the
'personal' factors, selfhood, human rights, etc...   the consequence of
addressing these issues leads to solutions that address broader
implications such as corruption[8], and in-turn support for human rights,
productivity; and therein, as we strive to transformationally improve the
lives of billions of people & achieve the SDGs, better investment
management, probity, etc...   particularly important when seeking to
achieve the goals of UN related projects such as
https://projectconnect.unicef.org/map  - that's about schools, not
club-houses for warlords...  although, depending on what the investment
purpose is overall, i guess..

my focus - peace infrastructure...  and the tooling, imo, to best support
it.

SO, one option might be to figure out how to define how this could be done
via a local agent - which would implicitly denote a process that would also
seek to address various implications; including but not limited to, privacy
of thought-ware...


*Browser Implementation*
So, in an effort to seek to address some of the above issues with something
- a reference implementation project - or similar...  noting, i'm still
very much looking forward to hearing feedback from others in the group...

I've been working on a web-extension based project, whereby the 'social
web' gets integrated into the browser using the web-extension framework.
The back-end data-store target has been solid[9], and the intention has
been to build something out that can act to better support growth of
conversations and related works, about various topics, including the human
centric AI works pipeline, et.al...

The focus at this stage, starts with human beings in their private capacity
- thereby able to be extended upon that foundational basis; inclusive, to
consideration for the need to support guardianship relations.

note, i understand this sort of thing may not be wanted by all - but rather
- so long as it improves support for human rights, rule of law, etc.  imo,
it should be allowed.

Thereafter,a few concepts i've been baking into it.

a. Values Credentials
The values credentials, are illustrated previously by my note about the
human rights instruments being augmented to support the use of the
underlying concepts as undertakings between the parties; which can then be
used, to support the instantiation of relationships via solid based
methodologies; and in-turn also, the management of them (ie: the agreements
between parties, can be updated / changed, etc.).

b. the ability to AUTH a domain name as an alternative to a FOAF URI.

c. reworking the social fabric: philosophical engineering / knowledge
engineering

So, traditionally - people have accounts on various platforms that segment
their social graph via those platforms.  but in a Human Centric (ai)
social-web implementation, that all changes.

there's two primary modalities,

(0. sleep).
1. recreation
2. work

Then, there's - in-effect, a life-log and/or web-log, which could (should?)
have support for website owners - to improve support for things like
'bookmarks', etc...

or the way a persons 'diary' or 'directory' functions in ways that's
broader than 'friends' or similar.

So, after some work on it, I've ended-up going back to the language's
considerations, noted previously as 'sense'...


*'sense' and HDF5[10]*
Noting, i'm not the best coder in the world...  which is not to say, i'm
not 'technical'....  nonetheless,

As i was implementing this web-extension (firefox target), i got to this
juncture where i was thinking about how to organise social relations,
alongside other underlying - ontologically associated -requirements, imo.
Which led back to the prior work on trying to figure out how to implement
natural languages, in full, spatio-temporally, as a means to provide
foundational support for the informatics environment - and fundamentally,
human centric AI / personal (and private) AI agents, on consumer devices.

My Concern about not doing it, is that any app would be required to
constantly make queries to support a persons 'thoughtware', which doesn't
sound like a good goal to pursue, at least, not to me.

coupled to this requirement for (AI) language support, is the need to
support declarative logic, and vector structures.  I'm not sure how this
can all be done in a browser for this future type of human centric ai -
"web x.x", that is different to web3 or similar.   so, my thinking for
implementing a POC was to look at prolog & julia, and I've then looked for
a data-structure format to support both semantics and vector related
data-structures, which yielded discovery of HDF5, that appears to be
equipped to support n-dimensional data-structures, in a format that doesn't
require the entire file (or files) to be loaded into memory.

So, there's existing resources like;
Babelnet, Cyc, FrameNet, Framester, Lemon, Linguist List, Suggested Upper
Merged Ontology (SUMO), WordNet, etc.

Some are in RDF, and there are other RDF data-sources; however, the level
to which existing systems provide Etymology is unclear; and not all
languages are presently supported, Some were never defined by cultures that
used books.

noting also; the meaning of words change, based on place, time and -
in-effect, context.

So, the underlying 'context' and purpose; isn't simply about ensuring
natural persons can understand one-another, but broadly also, that the
inferencing and foundational scaffolding for broader 'human centric ai'
systems, are empowered with context - declaratively, but ensuring natural
persons can define the intended meaning of words and thoughtware[12]
systems defined, prior to those data-sources being consumed by software
agents.

in-turn,

This is hoped to form some of the foundational pieces needed to support the
construction of personal ontologies that are able to be made interoperable,
whilst perhaps not part of how 'solid' works today.

historically, the intended designs - relating to old work, that led to
webid, rww, web-payments, credentials, etc....  intended to support a
cryptographic fabric; to ensure the outcome, 'human centric', supported
root-cause analysis and evaluation / support, for rule of law - now so long
after magna carta, as the digital transformation discussions are underway
via UN & other forums, the means for people[13] to define terms between
one-another online, as to form binding agreements, basically doesn't exist.
 no negotiation required, just comply or see ya later... when a person
wants to regain access to their electronic records...

The 'verifiable claims' people need, to have a voice & to look to the law -
as a source of peace[14]...

so, whilst understanding that 'AI', as is now commonly known to be like
'GPT' models, rather than semantics, (personal) ontology, or stuff like
HTM[15]; IMO, there is potential for significant performance / qualia[16]
improvements; yet, the cryptographic fabric, may be important for safety.

I also wonder whether the discovery of HTF[11] invokes some, somewhat
miraculous, opportunities to ensure support for portability of AI models /
systems, as to extend works on 'safety protocols', as i've hoped this group
would lead to forming useful solutions to address, as to ensure against
potential problems like mindware (personal (ontology) AI agent) platform
lock-ins, etc...

Therein; as much as 'language' models should, seemingly, be made able to be
distributed in HDF files, I imagine the same could be said for various
other forms of 'ai' libraries, that could be invoked on an as-needed basis,
by a local and/or personal AI agent; as well as similar considerations for
group AI agents systems.

In anycase,

The intended outcome would be, that whilst the ability to curate and
organise a persons 'thoughtware' landscape becomes ontologically defined
via natural language resources, the derivatives can then in-turn be
uploaded to a users 'solid pod', in terms of defining some sort of
interoperability target / standard - for 'cloud storage'[3], and in-turn
also - safety protocol outcomes, such as addressing the portability issue;
whilst perhaps, not entirely..

back in 2017, i wrote this presentation about 'fixing fake news'[17].   the
combination of implementing a 'sense' related system, with a
browser-extension; that supports both social engagement, annotations, more
advanced 'bookmarks', and personal 'knowledge' and activity log(s) - data
management (both personal & work); should help to illustrate some of the
underlying considerations about aspects of importance when seeking to build
technology to 'sign souls'.  Personally, I prefer the idea of using
category theory to better categorise content, rather than seeking to
address and/or find the 'fake book' section in a library...  afaik, the
closest thing to a 'fake books' i know of, is a 'bath book' for toddlers,
or similar, but they still, mostly, have some sort of story on it...  but
overall, there's differences between different people, different groups...
 for some, the term 'thongs' means underwear, for others it means footwear,
and the term 'gay' has different meanings depending on the temporal
context. My work on evaluating the implications broadly, have led me to
gain a better understanding also, of the complexities that evolve both
historically, in-terms of defining 'objective realities' as well as the
n-dimensional aspects associated to systems that might act to better
support (aka 'fit for purpose') inter-agent, social systems - made
interoperable with social machines[18]; whereby, Ensuring 'sense making',
across the multiple dimensions associatively, to support sense-making, in a
'fit for purpose' manner - seems like a matter of importance for human
rights...

and perhaps,

HDF5[19] is a good file-format to look to employ, for import/export, and
sharing of informatics catalogues of various sorts, for Human Centric AI
related applications.  Whilst keeping in mind, the requirements to support
health-care[20] - for all members of our human family, irrespective of the
jobs they may have, or that their loved ones may have... dignity, safety,
etc..  important...

*Spatio-Temporal encoding*
Spatial and temporal factors, are essential for many reasons, to support
context, and is instrumentally part of the designs i'm working on, to
improve support for 'sense making' and in-turn also, human rights...

*In Conclusion; *

If anyone is interested in helping with the 'browser extension' research
project, let me know...  I was hopeful that i / we, could define an
installable reference implementation that could then be employed to improve
our capacity to collaboratively / cooperatively, advance these works,
alongside more broadly otherwise - acting as a useful tool for people using
it, whilst using it to figure out how to make it better and what the basic
requirements are, as to further inform the scope of works for this w3c
human centric ai groups work...

Indeed also, getting some feedback to better understand whether these sorts
of works are going in a direction that is considered desirable by other
group members, is also a fairly critical purpose for my illustrations, per
above...

Perhaps there needs to be a multi-modal function / strategy, or different
categories, or better alternatives, etc...

I look forward to some productive feedback...

Finally also,

it would be good to organise a group zoom.  some feedback as to what
time/days best suit people, would be helpful to define this.  If others are
interested in presenting, this is also welcomed.  I'm not sure how many
presently know and/or know of one-another, making progress in this area
would also be welcomed...

Hope this helps;

Kind Regards,


Timothy Holborn
http://www.humancentricai.org/
https://www.linkedin.com/in/ubiquitous/

*LINKS:*
[1]
https://www.un.org/techenvoy/global-digital-compact/intergovernmental-process
[2] https://twitter.com/DameWendyDBE/status/1172470883610431489
[3] https://www.w3.org/DesignIssues/CloudStorage.html
[4] https://en.wikipedia.org/wiki/Internet_governance
[5]
https://en.wikipedia.org/wiki/Internet_governance#/media/File:Who-Runs-the-Internet-graphic.png
[6] https://en.wikipedia.org/wiki/Objective_structured_clinical_examination
[7] https://data.worldbank.org/indicator/SH.MED.PHYS.ZS
[8] https://press.un.org/en/2018/sc13493.doc.htm
[9] https://www.w3.org/community/solid/
[10] https://en.wikipedia.org/wiki/Hierarchical_Data_Format
[11] https://github.com/HDFGroup/hdf5
[12]
https://cdn.knightlab.com/libs/timeline3/latest/embed/index.html?source=1r-bo83ImIEjSCmOFFMcT7F79OnCHDOGdkC_g9bOVFZg&font=Default&lang=en&hash_bookmark=true&initial_zoom=4&height=750#event-consciousness-qm-ai-studies-video-edition
[13]
https://www.ohchr.org/en/human-rights/universal-declaration/translations/english
[14] https://vimeo.com/30416090
[15] https://en.wikipedia.org/wiki/Hierarchical_temporal_memory
[16] https://en.wikipedia.org/wiki/Qualia
[17]
https://drive.google.com/file/d/1fvrYlTYUCYQMjW4pTel6QOIvJ03GtQT7/view?usp=sharing
[18] https://link.springer.com/book/10.1007/978-1-4842-1156-4
[19] https://github.com/search?q=hdf5
[20] https://ec.europa.eu/commission/presscorner/detail/en/ip_23_3043