I try to be careful about my allusions to "openness". I attribute (perhaps
wrongly) the openness of science to Critical Rationalism (Popper, but better described by
David Miller). Good (and bad) ideas can come from *anywhere*. Even those miracle people
like FGJ Perey can come up with bad ideas. My (false) dichotomy between nonsense and
abductive triggers might be problematic. But that's just a distraction. The real point is
about the interstitial spaces *between* models, not the models or the ground they cover.
I would never ask whether James and/or Husserl are correct. But asking whether they're
talking about the same thing is a good question.
The backend of the LLMs seem like mere validation, to me. What's more important is whether the
weights and structure of the underlying system span the space of the (natural) language. Let James
and Husserl be language models, J and H, themselves. The way J would determine if H is talking
about the same thing J is talking about is by talking. When the language expressed by H is enough
*like* the language expressed by J, then J will say "yes". If the expressions stably
remain dissimilar, or explode in dissimilarity, then J will say "no".
Objectively, if J and H are capable of "online learning", then we might be able
to do some kind of similarity measure of their language before they interacted. Then let
them yap with each other for, say, 1000 iterations. And measure the [dis]similarity
between their expressions at time 0 and their post-interactive expressions. Maybe H
brainwashes J such that d(H0,H1) = 0, but d(J0,J1) = d(J0,H0) ... or vice versa.
On 6/5/23 10:45, Steve Smith wrote:
If/when/as AI (such a broad term, no?) can be used in the mode you describe here somewhat
transparently I would likely be open to an "augmented intuition" mode of use....
and as a point of gratuitous contention, how *does* one tell the difference between "stupid
nonsense" and "an abductive candidate for experimental research"? Is there truly a
qualitative difference (in the world) or is that an artifact of our own judgement(s) based on some
quantitative threshold(s)?
Your description of "T" Truth as a spanning kernel for a plurality of theories and models feels quite apt and
the way I took the name of the Docuseries "Closer to Truth", very assiduously avoiding the specific "the
Truth"... and implying an "asymptotic" approach not a collision course.
As I look at the (near) decomposable systems and map it onto (near) spanning trees within
process-relation networks of those systems I imagine these LLM training exercises
building/finding highly connected clusters (like ganglia in vertebrate neural systems)
which fundamentally reflect what KellyAnne Conway so naively claimed as "alternate
realities".
If there is a singular capital T Truth (or capital R Reality) then it is
probably at most apprehended by finite beings (who have not achieved Satori,
nod to DaveW) as the superposition of many sub-complete T' (or R') descriptions?
- Steve
On 6/5/23 10:18 AM, glen wrote:
But this misses the point, I think. And, in fact, I think it's a mistake to focus too
much on (natural) language models at all, even for things that *seem* to be all about
language, like philosophy. I'm most interested in the concept of an embedding
<https://en.wikipedia.org/wiki/Embedded>. Ideally, I'd like to be able to query
a modeling system (e.g. decoding/encoding transformers) for a (vector) space that
(accurately) encodes *both* James and Husserl. Then that would help satisfy Marcus'
and EricS' task to see if there are gaps between them, or not.
The problem has nothing to do, really, with how much one might have to
read/think in order to understand anything. Understanding is a delusion. What
matters is the differences and similarities between any 2 or more things
(processes, devices, systems, whatever 'thing' might mean).
These automatic modelers (like the transformers) might help us do that. As for some kind
of "ground truth", something that might provide a foundation like the
physicists seem to think they have, if our automatic modeling device is capable of
embedding all (or most) of all the models surrounding us (over time, space, and
individual theorists or collections of theorists), then we can experimentally test for
kernels/bases that can span *most* of those theories/models. If such a kernel exists,
then it is a candidate for the capital T truth, and any theory/model that is not spanned
by that kernel is either stupid nonsense or an abductive candidate for experimental
research.
On 6/4/23 18:48, Frank Wimberly wrote:
As one of the few, if not the only, person who has been a full time employee of a philosophy
department for multiple years, I am quick to defend my former colleagues. Read "Actual
Causation and Thought Experiments" by Glymour and Wimberly in J. K. Campbell, M.
O'Rourke & H. S. Silverstein (eds.), Causation and Explanation. MIT Press
You don't have to read thousands, or even hundreds, of pages to be able to grok
that paper.
---
Frank C. Wimberly
140 Calle Ojo Feliz,
Santa Fe, NM 87505
505 670-9918
Santa Fe, NM
On Sun, Jun 4, 2023, 7:30 PM David Eric Smith <desm...@santafe.edu
<mailto:desm...@santafe.edu>> wrote:
So there’s a rather concrete way in which one can imagine ChatGPT’s being
particularly useful as a time-saver.
I have heard it said (and find it persuasive), that philosophy is different
from physics because what philosophers want to do and settle for being is
different from that for physicists.
A physicist can pick up F = ma and start from there to get something done.
Each philosopher is, in a sense, a new beginning of the universe, and you
are expected to read thousands of pages of his composition to be permitted to
engage with him. That is a good barrier to exclude pretty-much-everbody from
most conversations.
But there are specific topics on which engaging with this group is a game
of whack-a-mole, and it would be _so_ satisfying to catch that damned mole far
enough out of the hole to pin him down to the board for once.
It is on this point:
Summarizing what, as Marcus rightly says, as been repeated 10^n times
before, CGTP quotes:
At the core of radical empiricism is the concept of "pure experience."
According to James, pure experience refers to the immediate, unmediated apprehension of
reality, devoid of any conceptual or interpretative filters. It involves experiencing the
world as it is, without imposing preconceived notions or theories onto the experience.
What the HELL does anyone think this is supposed to refer to? I am not
asking whether it actually does refer to anything, but rather what anyone
believes he is saying by it.
And I can ask that in a rather concrete way. Were James to engage with
Husserl, would he claim that the access to the “immediate apprehension” is by
way of the same portal as Husserl’s epoche?
I ask because they set themselves up to make a particular style of
assertion.
By analogy, we have seen that human bodies can do things like Amanars and
any of the 4 Bileses (which should have been 5, and would have been were it not
for COVID). But that doesn’t mean every human body can do any of them. There
is rather a lot of specific training that goes into becoming one of the bodies
that can do any of this.
The various “internal” experience-focused philosophers present these things
as doable, but technical and particular and requiring training.
But if you then ask what that is about, you get either a demand to follow
several thousand pages in each person’s formulation, or the kind of cloudy
motivational life-coach speech that almost all of the CGPT summary is composed
of. (Reminds me of something I once heard said of chimp speech: if you aren’t
there working with them, you cannot anticipate how mind-numbingly repetitive it
is).
So rather than asking “what it is” (the skill or whatever), I can ask “If
they were arguing with each other, would they even assert to each other, each
with his supposed privileged appreciation of the mysteries, assert or deny that
they are referring to the same thing.
This might allow us to not have to approach the full body of philosophical
literature as if each corpus were Sui generis.
Eric
On Jun 5, 2023, at 2:43 AM, Jochen Fromm <j...@cas-group.net
<mailto:j...@cas-group.net>> wrote:
ChatGPT now allows sharing conversations. I've asked it about William James book
"Essays in Radical Empiricism"
https://chat.openai.com/share/375aef4e-a8d6-467e-8061-bd85b341c46b
<https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fchat.openai.com%2fshare%2f375aef4e-a8d6-467e-8061-bd85b341c46b&c=E,1,SrWav4ypspeJiXxANsU84IqWFKy5OPWIx-qHp0YLHpEHLinoe3Q3aAeuo_0eErOe6fnJYosh3T6fflwMl7CsxV2wKAIIwCbBlleeoZM8db1fEE4,&typo=1&ancr_add=1>
--
ꙮ Mɥǝu ǝlǝdɥɐuʇs ɟᴉƃɥʇ' ʇɥǝ ƃɹɐss snɟɟǝɹs˙ ꙮ
-. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. .
FRIAM Applied Complexity Group listserv
Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom
https://bit.ly/virtualfriam
to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
FRIAM-COMIC http://friam-comic.blogspot.com/
archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/
1/2003 thru 6/2021 http://friam.383.s1.nabble.com/