Yes I agree, the epistemological challenges here are… incredibly non-trivial. When we start talking about arbiting truth vs misinformation, and building systems around this, it immediately turns to philosophical questions around what types of knowledge are subjective vs objective, and what it even means to “know" something. I often worry that the builders of these types of systems aren’t thinking about these types of things.

Also, when it comes to chatgpt, it’s always best to keep in mind the goal was to emulate a human conversation, with no guarantees (or even explicit design goals towards) useful or truthful information. See also one of my longtime favorite xkcds: https://xkcd.com/810/

On Jan 4, 2024, at 16:18, rjmcmahon via Nnagain <nnagain@lists.bufferbloat.net> wrote:

Hmm, this seems non trivial. The epistemology that I try to use is based on the scientific method but I also use a lot of belief that others follow this method too.

Also, with the current trends in AI, it seems to me we're mostly creating ant mills vs knowledge and discovery.

https://youtu.be/LEKwQxO4EZU?feature=shared

I asked ChatGPT if it behaved like an ant mill. It responded no. I then asked it to compare itself with and ant mill and basically said it behaved exactly like an ant mill.

Then I asked it some questions about Elon Musk and it always prefaced everything with, "Elon is a God."

I stopped using it after that. No knowledge to be found.

Bob
Dear NNAgain’ers,
Today on a different listserv, I joined a discussion on what I sense
will be a pressing issue across multiple sectors in 2024. I recognize
this is not NN-related and so if it isn’t of interest, I apologize
in advance. However as most of us have technology background here, my
sense is we generally have a better sense of the looming issue than
non-technical folks at the moment. Below I outline some of the
contours of the evolving problem space, and invite each of you to
share your thoughts as I sense the diversity of perspectives here
might help with brainstorming potential solutions necessary for civil
societies to continue:
Premise: We are at the precipice of an extended era where
inauthenticity vs. authenticity will be difficult to discern, that
that involves multiple forms of content including biometrics and more.
In isolated pockets, governments are becoming aware of this - however
it’s going to be really difficult for pluralistic societies like the
U.S. where any of the Estates that traditionally would have a role to
play in verifying the authentic vs. inauthentic nature of something
have had public trust in them as arbiters eroding. And it doesn’t
help that both politics and advertisement rely on presenting things as
100% authentic when they’re often only somewhat so (or, to be more
generous, mix facts with lots of beliefs).
Not supporting autocracies, however they have a bit of a “home
field” advantage here because there is only one singular narrative -
and anyone who questions it can be fired/isolated,
imprisoned/disappeared, or killed/executed. Tools of such regimes, to
include filtering, censorship, and repression - will be used to ensure
only one narrative (authentic or not, mostly likely the latter) is
seen by a majority of their population. Pluralistic societies will
have it much harder, and the last ten years will pale in comparison to
the challenges of sensemaking in a world flooded by both media and
mediums of questionable authenticity.
Back in 2019-2020, I did my darnest to connect Pablo and an additional
People-Centered Internet expert with Salesforce that has a lot of CRM
data with the proposal that SF could provide a feature where, as part
of the CRM, “out of band” questions could be included to do some
sort of additional level of trust that the entity on the other end was
who they claimed to be. Unfortunately that pitch was overshadowed by
larger concerns that SF’s software, give some of its features, could
be misused in ways not intended by them (think about ways akin to
Cambridge Analytica) and they were trying to figure out how they could
incorporate features to prevent actors from misusing/abusing their
software in ways not intended by them as a company.
2024 is going to be hard. Manipulation of what people appear to see,
hear, sense - and thus know - is becoming sadly easier.
Meanwhile understanding of the importance of triangulation,
triangulation, triangulation from different perspective to discern
authenticity vs. inauthenticity remains time-consuming and hard.
Perhaps we need to consider standing up private sector Dun &
Bradstreet-like entities for identity and other important adjudicatory
functions - however that doesn’t immediately solve the issue of how
to help the public in a would experiencing a flood of questionable
content, information, and identities? And who “watches” the
adjudicators?
David Bray, PhD Principal, LeadDoAdapt Ventures, Inc. [1]
Loomis Innovation Council Co-Chair [2] & Distinguished Fellow
Henry S. Stimson Center [3], Business Executives for National Security
[4]
Links:
------
[1] https://www.leaddoadapt.com/
[2] https://napawash.org/fellow/305629
[3] https://www.stimson.org/ppl/david-bray/
[4] https://bens.org/people/dr-david-bray/
_______________________________________________
Nnagain mailing list
Nnagain@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/nnagain
_______________________________________________
Nnagain mailing list
Nnagain@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/nnagain