Re: Interest in a zoom meeting on safe AGI?

HI Dave,

probably more text than is helpful; but there's a few links in it..  etc..

When I get to writing some code examples, I'll let you know. but that also
seems foolish, as you'd be much better at it than me... Indeed, I'd also
welcome your help, but I think we're both focused on different things;
that, I hope, should be complimentary at least in some areas.

FWIW: i think this clip, is fairly awesome; https://youtu.be/ZYPjXz1MVv0  |
i also found https://www.youtube.com/@models-of-consciousness/videos
recently, maybe there's something in there you'll find useful / helpful...

Also, I've had a prosthetic eye since I was about 1.  the amount of control
that is sought to be made a mandate upon all members of our human family,
in relation to the operational rules sought to be placed upon us, in
relation to the operation of our thoughtware; indicates to me that there's
some very concerning issues, and I think alternatives are important.

“the distinction between reality and our knowledge of reality, between
reality and information, cannot be made” Anton Zeilinger   (
https://www.nature.com/articles/438743a )

I don't think we understand how the 'mind' works, consciousness, et.al..  i
therefore consider the most important job, is to make sure the work to
produce good enough instrumentation to find out - is done,
honourably. Obviously therein, we're both here...  very few are, and the
terms are often punitive; but that all depends upon values, etc...

best wishes,

Tim.h.

On Thu, 9 Nov 2023 at 00:38, Dave Raggett <dsr@w3.org> wrote:

>
>
> On 8 Nov 2023, at 11:59, Timothy Holborn <timothy.holborn@gmail.com>
> wrote:
>
> I think there needs to be an update from foaf. Was hoping for natural
> language ontologies.
>
>
> The use of LLMs for combatting disinformation and inflammatory posts isn’t
> related to RDF and associated ontologies. It is more like having a tireless
> friend to vet what you see and send on social media. This friend can block
> out objectionable content or at the very least tell you what’s wrong with
> it.
>

would be better if the 'tireless friend' operated locally, rather than
beaming every keystroke (thin client); but also, its very 'web 2.0' -- not
that the concept of 'web 3.0' is very comprehensible anymore....

if people are operating their own social domain ('social web', 'social
layer to 'the web'') ; (ie: cimba) - kinda works differently - yet
certainly, there are still 'safety protocols' needed...


>
> Social media companies will be required to take reasonable steps to
> enforce their rules and conditions in respect to posts.  This will require
> some general agreement on what constitutes disinformation and inflammatory
> content.  Neural networks are excellent when it comes to working with fuzzy
> context sensitive concepts based upon foundation models and reinforcement
> learning with human feedback, including assessment against loosely worded
> principles.
>

I do appreciate the need for it - particularly in the context you speak
of...  but the context i'm looking at is moveover that 'web of data' level
concept.

Therein;  words have different meanings, attached to different places and
times...  whether it be the use of a term 'gay' in a paper from the 1920s
or the use of the term 'thongs' from an Australian Context (meaning -
'flip-flops'?  footware)...

By augmenting the way content is stored to provide expressive semantics,
context...  the means to then send it through LLM programs & other ML / DL
/ etc..  is an extension to the foundational need to support personal
ontology, personal cryptography, etc.  LLMs processing data that is of poor
hygiene, lacks context or has massive 'gaps' with regard to the data that's
available in relation to the topic, subject, etc..  will still end-up with
different results.  objective is good signal to noise ratio.  BUT THIS IS
DIFFERENT....

I understand various groups in the US find it hard to consider the merits
of why at least some people want to have some level of 'selfhood' support,
in relation to their 'digital prosthetic' of self; human agency, and indeed
also, the implications relating to the development of children; who, in
some parts of the world, increasingly see adults who are not paid to care
for them, less and less...  due to pressures & broader priorities.

AND; I appreciate how some people may like to block a former friend, as
though they never existed at all - but this isn't really compatible with
our mindware, it's a means to support older models; that I strongly believe
will be disrupted, for various reasons, some pragmatic; others moral.
Indeed, it was jobs as he launched NeXT that spoke about 'inter-personal
computing'...  So, whilst this is likely a massive challenge for many w3c
stakeholders; something that's difficult to get their head / business model
around;

https://www.youtube.com/watch?v=B635wcdr6-w

https://www.youtube.com/watch?v=4QTAtFaIiyc

A priority for me, was to ensure technology provided the means to support
human rights, by ensuring people could take electronic evidence - of their
lives, into a court of law to have problems sorted out.  Governments
seemingly took-over, and built systems that protected particular government
programs that broke the law, and people...

https://en.wikipedia.org/wiki/Robodebt_scheme

this didn't work out for them either:
https://www.theguardian.com/australia-news/2023/nov/06/scam-in-a-box-mygov-suspends-thousands-of-accounts-linked-to-dark-web-kits

information, is information; fiction, non-fiction, claim, etc. like the
different sections, genres, etc. in a library.   The foundational
considerations relate to our values, such as those described by human
rights instruments alongside others, such as international humanitarian
law, various forms of community values instruments, etc...

When it was discovered that the world was not flat; i'm sure the claim was
considered both 'false' and 'inflammatory', but it was also correct - which
took time to figure out.    Increasingly nowadays, the initial wrong
committed by an origin record of a longer-term problem; is not as
cumulatively bad, as the consequential acts made by many others until such
a time that - a policy changes...  often without regard for who benefited,
and who wore the cost.  But, this is a major productivity issue that
disaffects insurance, financial markets and all sorts of other social
faculty.

https://github.com/WebCivics/docs.humancentricai.org/blob/main/Attachments/Diagram16.jpg
https://github.com/WebCivics/docs.humancentricai.org/blob/main/Attachments/Diagram14.jpg

https://github.com/WebCivics/docs.humancentricai.org/blob/main/Attachments/Diagram12.jpg
https://github.com/WebCivics/docs.humancentricai.org/blob/main/Attachments/Diagram11.jpg

so, from the 'consciousness' market - there's an investment cost, into
producing information systems or knowledge systems; but these are different
forms of design practice, requiring different structures - IMHO...  end-up
with tech debt (and worse) if it's later figured out, that the wrong path
was elected - irrespective of who foots the bill...

https://github.com/WebCivics/docs.humancentricai.org/blob/main/Attachments/Diagram8.jpg

The most recent 'safety protocols' doc is:
https://docs.google.com/document/d/16rrLkhxzuSMg1TXVHnzTOLtMUsRZB4zKpKnKFnjtcao/edit#heading=h.sl7he2jnzuo6

The 'Social Attack Vectors' document is:
https://docs.google.com/document/d/19s7HbEUbt4_mKOvq9xmZhi-305zEMnDINekiJOpECTc/edit#heading=h.xve4xlnszz4m

IMO: there is a means to define 'logical programming' methods that can flag
certain behaviours, no-matter how sophisticated; in a manner that's better
than the capabilities of https://www.maltego.com/  or similar.

but these systems designs are different; to the large-scale alternatives,
that seemingly require everything to be sent 'to the cloud' - so, it's a
bit like evaluating the needs for personal computers vs. those of
mainframes...  it's just a different age, some love microsoft, but aren't
we all happy that it was allowed - to produce linux...

Further - these ecosystems i'm working to define, will most certainly be
able to be interfaced with cloud AI APIs..  it's just not required all the
time, as a mandate of sorts...  the way i see it, people start by defining
the values that they share, then negotiate the terms that they want to
employ to define the 'shared values' established at the commencement of
their relationship;
https://github.com/WebCivics/SocialWeb-WebExtensionDev-v4/blob/main/valuecredentials/flow.md

people have domain names; and can delegate email alias to represent the
relationships, as they own the domain...

There's lots of credentials, lots of verifiable claims and a bunch of AI
tooling, some of which could be built into the browser (ie: social web) but
- regardless, alot could probably be done via a local virtuoso install;
even though, my preference is for a RUST wrapped, prolog/julia local
web-server, plugged into wireguard using WebID-TLS to connect with a host;
that helps with KYC/AML, IPv6 subnets,

then, the other thing, from a cyber-security perspective (was watching the
DNS session earlier).

imagine this;

1. user has Domain Name & IPv6 (WebID-RSA or similar in domain records)
2. Connection peers WebID-TLS
3. users 'web-of-data' use OIDC to specific provider
4. bunch of verifiable claims / credentials from 3rd parties (able to be
used with local Human Centric AI agent)
5. HTTP signatures / various document signatures
6. decentralised topology; with decentralised 'permissive commons' methods
for supporting distribution of various types of commons (could be personal,
could be universal, like language for multi-agent support (inc. human
centric ai agent, context based assets, etc).

so, overall, ends-up with a bunch of different cryptographic tooling; which
creates what I think of as a 'social cryptography fabric'...

so, imo, the architecture is different to web 2.0 & Web3, etc...

Per my ISOC Pitch: https://www.youtube.com/watch?v=VgenJKdd9CM  - safety
protocols are optional (notwithstanding what otherwise exists on the
network...); but therein, protocols works like NymTech seek to offer
rock-solid security, which is important due to corruption issues which are
really the biggest enemy of all; but the balance is, that people could run
different safety protocols, they should highlight that they're running a
'child safety' protocol or display their police check or other attributes,
in relation to the context of that relationship - and, overall - this helps
address the 'objectionable content' issue - which is a really hard topic -
if done properly....  whilst some prefer a few people in the world to
define the answer to all things, for all around the world; this isn't the
case for everyone..  so, i think there's demand for an alternative.

around the world...

but most of all, there's a massive need to create systems for places, with
people whose languages are not supported by the internet today; that may
not have anything to do with books; and somehow, we need to figure out how
to use technology to help people, forge a medicinal earth environment
around them, to gain knowledge, share knowledge, deliver SDGs, obliterate
'digital slavery' whilst being cautious not to introduce it into places
where it is not already a problem; and, advance, as a species...

From early this year - before i established the Human Centric AI CG:
https://soundcloud.com/ubiquitous-au/my-question-human-rights-instruments-for-digital-wallets-valuescredentials

https://www.linkedin.com/feed/update/urn:li:activity:7125747029506437120/

some people find it objectionable that the consequence of their actions may
lead to court actions, to determine - peacefully - an outcome for a
dispute, and consideration about the costs incurred, etc..

other people, seek to look to the law, as a source of peace
https://vimeo.com/30416090

People design technology, as artists, expressing - like an extension of
their soul, their values, what it is that they consider to be important;
and what is ok, to be set aside...

the designs today, don't provide access to justice, particularly for the
poor; and whilst basic tooling has been produced, the volume of work that
now needs to be done after the 'we're doing it all already' crowd were
provided their opportunity to go do it all - shows, that in fact, they
weren't doing it all..  and people are still suffering.


> Dave Raggett <dsr@w3.org>
>
>
not all friends are friends; people should be able to define ontologically
the common-interests they have with others, which helps them manage how
they interact with others, socially - as an extension of their browser;
 and some, even loved ones whose voices either party have never heard; may
well be working, on stuff, to ensure a future level of safety that did not
exist earlier.  people don't take their kids to war... but many will attack
them, particularly in relation to corruption issues, where reprobates find
means to make a dollar - due to the lack of accountability systems...
particularly hard, when associated with government funding...  So, if
you've got means to address those issues faster than I can - please, let me
know... but for now, I understand I'm required to just 'suck it up', so
those involved in harming people, don't get upset... After all, without
infrastructure to provide comprehensive evidence of facts to a court of
law; these people are continually empowered to make new attacks, as there's
no real consequence - other than the profit they make... it's fairly
sickening really, the illustration of priorities - now demonstrated...  I
honestly don't know how the 'we trust global platforms' model can possibly
address the problems, the digital slavery economics...

but, whilst there's many applications for these sorts of 'human centric' ai
agents / ecosystems; i do think they're quite different to platforms, etc.
 indeed, when i defined the notion for creds / rww - whilst the point also
related to software licenses (ie: the ability for people to own their
'inforgs'); it was intended to be different to the pre-existing models,
that were so entrenched that i couldn't get funding by VCs or Gov; as VCs
wanted to own all the data, and Gov wanted to ensure they could cover-up
their mistakes - even when, the victims were children...
https://www.w3.org/Search/Mail/Public/advanced_search?keywords=%22human+centric%22&index-grp=Public__FULL&type-index=&resultsperpage=100&sortby=date-asc


FWIW: oranges & sunshine is a good film: https://g.co/kgs/a45GDr  - seems,
they've just continually evolved the models over time, to seek to maintain
a supply of children...  it's reprehensible.  so, sure.  Some will want the
'epstein island' compatible infrastructure; others, want to be protected
from anyone who has anything to do with that very different ideology &
related world.

to deliver SDGs, which is for billions of people, hundreds of millions of
houses, unknown number of slums, people throughout the world, seeking
kindness - but, often in places that have major problems with corruption -
fairly sure, investment into corruption, isn't considered one of the
intents & purposes for SDG related works, noting UN has had problems:
https://www.youtube.com/@Andrewmmacleod/videos

and frankly also,

if there's insufficient support via english speaking 'west'; then, i don't
see why these sorts of works wouldn't end-up being advanced by CE MFGs &
packaged with their devices, probably at lower cost...  but if that does
happen, i'm also not sure what internet it would be connecting to...

such are my concerns about the implications of corruption...  said, to be
at least 5% of GDP: https://press.un.org/en/2018/sc13493.doc.htm  - but at
an unknown cost in terms of the associated volume of CO2...  or ESG issues,
etc..

but overall, i do hear alot about experts... apparently there's alot of
them, which is interesting because they seemingly fail to deliver; and
then, there's people such as yourself who i consider to be remarkable, but
when trying to figure out how to address issues - seems humanity at large
are MIA.

I am hopeful that the ISOC thing MAY help to address these issues; and, i
think the likely outcome for Human Centric AI, is that this fight over what
the term means - will lead to having different categories of the broader
concept, as to illustrate the feature set of different types of modal
structures...  some, will want to call gov run thin-client systems, 'human
centric'..  but that's not what i'm talking about...  because magna carta,
was NEGOTIATED in a church in...  holborn...  and their solutions, just
express, now almost a thousand years later, moral poverty...  but
apparently that's all changing now...  I haven't seen the funding yet,
whilst I got left with the bill.

best wishes,

tim.H