<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<p>Cody/Glen -</p>
<p>I like this conceit of dreaming as disinhibited/discursive
next-token generation. <br>
</p>
<p>As a sometimes *lucid* dreamer myself, I feel as if that is what
is very close to what is going on for me... my best lucid
dreaming is hypnapompic... happening as I awake or as I drift in
and out after a good sleep (including various fever dreams such as
an overheated nap on a summer afternoon). Hypnagogic also, but not
so much... it seems to be a difference in tuning/calibration/angle
of entry vs angle of exit?<br>
</p>
<p>The *lucidity* of my dreams seem to stem from some kind of split
attention or an pseudo-adversarial exchange between a *conscious*
and an *sub*-conscious component of "self". The mode in which I
engage in my lucid dreaming experience is usually somewhat
active... "guiding" the dream-wander willfully but always
gently... experience tells me that deliberate lucid dreaming is a
bit like catching dust-motes or herding cats... too much intention
blows it up and shuts it down or ruins it qualitatively.
"directed disinhibition" might be a fair description of my best
method. Psychedelic (or opiod?) tripping may be similar for some,
my experience with such is so acutely limited that I'm just
guessing from anecdotal references.</p>
<p>One of my early experiences with chatGPT was to try to co-write
short fiction with it, based on some simple tech or
socio-political tweak. The final result was rather unsatisfying
but Glen's implication that an API interface might provide more
degrees of freedom than merely web-whispering at it. I hadn't
thought of it before this thread came up, but the *problems* I had
might have been in the category of *trying too hard*... I was
probably trying to force chatGPT to be more of a writing
assistant than a collaborative dreamer. Naturally I begin those
story-weavings with a specific idea of what it is about and where
it is going. Perhaps I can re-engage with something closer to
Glen's prescription?<br>
</p>
<p>Jason Yunkaporta in his book <a moz-do-not-send="true"
href="https://birchbarkbooks.com/products/sand-talk">Sand Talk</a>
invoked a (n Aboriginal) term for doing something like this
intentionally with other people and calls it "Yarning"....
somewhat discursive from the usual description of Aboriginal
Dreaming is my own connotative understanding of it as "wandering
about in the (larger) adjacent possible". Yarning seems to be a
collaborative, directed version of this? The question of "what
means adjacent?" is begged here methinks.<br>
</p>
<p>I have resisted/puzzled-over/enjoyed Glen's assertion/question
about "communication is an illusion" (probably mis-stating it
here?)... and wonder how that relates to this "dreaming"
business? It seems that what we nominally call (or intend to be)
communication is at best the weaving of yarns (and that is a good
thing)?<br>
</p>
<p>For the musically inclined: <a moz-do-not-send="true"
href="https://www.youtube.com/watch?v=89dGC8de0CA">Dream On</a>!
- Aerosmith or a <a moz-do-not-send="true"
href="https://www.youtube.com/watch?v=TYXYQRU2i7c">Cinematic
Slam version</a> (more to the spirit of adversarial
collaboration)...<br>
</p>
<p>For the western take on eastern mystics among us, <a
moz-do-not-send="true"
href="https://www.youtube.com/watch?v=wU0PYcCsL6o">Alan Watts</a>...
<br>
</p>
<p>Discursively yours,<br>
</p>
<p> - Steve<br>
</p>
<blockquote>
<p><br>
</p>
</blockquote>
<div class="moz-cite-prefix">On 9/8/23 8:03 AM, glen wrote:<br>
</div>
<blockquote type="cite"
cite="mid:00c88361-4a20-076f-3e80-93927d67c888@gmail.com">One of
the things we could easily try is cumulative, iterative prompting,
particularly with some of the lower scoring responses. Dreams are
nothing but lower scoring responses, right? While you're sleeping,
your evaluation/selection mechanism is inhibited, which allows you
to invest a little more in the total bullshit your own next-token
generator generates. So for, say ChatGPT to dream, it simply needs
instructions, including higher temperatures, to be less critical
of its own responses. It would be annoying to try to do it with
the web interface, but trivial to do with the API. The hidden
pre-prompt could be engineered such that the n+1 prompt is a
(algorithmic or random) composition of the, say, 10 responses to
the nth prompt. Etc. This would be akin to dreaming, I think. At
the end of however many iterations, you wake it up and write the
highest scoring result down in its dream journal.
<br>
<br>
Maybe I'll try that with Falcon. I can't divert my OAI budget to
it.
<br>
<br>
On 9/7/23 12:37, cody dooderson wrote:
<br>
<blockquote type="cite">I asked ChatGPT if it dreamed and it said
that it didn't. However, is adversarial training of neural
networks much different than dreaming?
<br>
<br>
A new class from MITX showed up in my email today. It is called
/Minds and Machines: An introduction to philosophy of mind,
exploring consciousness, reality, AI, and more. The most
in-depth philosophy course available online.
/<a class="moz-txt-link-freetext" href="https://mitxonline.mit.edu/courses/course-v1:MITxT+24.09x/">https://mitxonline.mit.edu/courses/course-v1:MITxT+24.09x/</a>
<a class="moz-txt-link-rfc2396E" href="https://mitxonline.mit.edu/courses/course-v1:MITxT+24.09x/"><https://mitxonline.mit.edu/courses/course-v1:MITxT+24.09x/></a>
<br>
It may help with this question.
<br>
/
<br>
/
<br>
_ Cody Smith _
<br>
<a class="moz-txt-link-abbreviated" href="mailto:cody@simtable.com">cody@simtable.com</a> <a class="moz-txt-link-rfc2396E" href="mailto:cody@simtable.com"><mailto:cody@simtable.com></a>
<br>
<br>
<br>
On Thu, Sep 7, 2023 at 12:25\u202fPM Steve Smith <<a class="moz-txt-link-abbreviated" href="mailto:sasmyth@swcp.com">sasmyth@swcp.com</a>
<a class="moz-txt-link-rfc2396E" href="mailto:sasmyth@swcp.com"><mailto:sasmyth@swcp.com></a>> wrote:
<br>
<br>
Great observations as usual Glen... I have lapsed into
*listening* to
<br>
almost all long-form writing, whether fiction or non....
and it
<br>
definitely distorts (torts?) my perception/conception of the
<br>
material/subject/message. A corollary to McLuhan's
Medium/Message
<br>
duality?
<br>
<br>
I find the "output" side to be more specific (or
conscious) for me
<br>
than the "input" side. Your point of cuneoform
<br>
sticks/quills/pencils/keyboard/gestural-interpreters being
part of our
<br>
extended phenotype is very apt as is the idea that (if I
understand your
<br>
intentions) it (intrinsically) effects our interoception and
<br>
inter-subjective realities.
<br>
<br>
I also appreciate your reflections on "mal" and "dis" which
I have lived
<br>
with all of my life... "judging" or "discriminating" in ways
which
<br>
themselves are "adaptive" for one suite of purposes but
perhaps
<br>
"mal"/"dis" for another suite. Having a vector or tensor
fitness
<br>
function with (arbitrary) signs on the elements doesn't
guarantee they
<br>
themselves are "fit" for what you think they are.
<br>
<br>
Do Androids dream of Electric Sheep? Do LLM's (or larger
adaptive
<br>
systems they are embedded in?) dream of the tensor fields
they are
<br>
embedded in or create or co-create with the fields of human
<br>
activity/history/knowledge/experience/future/manifesting-destiny
they
<br>
were designed to model/emulate/expose/facilitate/co-evolve
with?
<br>
<br>
I dunno, but it sure is a fascinating milieu to be surfing
through in
<br>
these auspicious days at the beginning (or end) of the
Anthropocene.
<br>
<br>
- Steve
<br>
<br>
On 9/7/23 1:01 PM, glen wrote:
<br>
> Both keyboards and pencils are part of our extended
phenotype and play
<br>
> (multiple) roles in interoception, including the
induction of
<br>
> inter-subjectivity. I've forgotten who it is, but
there's someone on
<br>
> this list who *listens* to our posts, rather than
reads them. I tried
<br>
> that with a blog post this morning during my mobility
routine:
<br>
>
<br>
>
<a class="moz-txt-link-freetext" href="https://www.emilkirkegaard.com/p/preferences-can-be-sick-mental-illness">https://www.emilkirkegaard.com/p/preferences-can-be-sick-mental-illness</a>
<a class="moz-txt-link-rfc2396E" href="https://www.emilkirkegaard.com/p/preferences-can-be-sick-mental-illness"><https://www.emilkirkegaard.com/p/preferences-can-be-sick-mental-illness></a>
<br>
>
<br>
> <tangent>
<br>
> Then because I had an allergic reaction to what I
heard, I *read* it
<br>
> later. Listening to it disgusted me. I came away
thinking this
<br>
> Kirkegaard dude's akin to a scientific racist ... or
maybe a
<br>
> eugenecist. I admit to being a fan of Thomas Szasz
back in the day. (A
<br>
> friend's mom actually dated him at some point ...
allegedly.) But at
<br>
> this point, I've been infected by the Woke Mind Virus;
and it's
<br>
> difficult to stomach phrases like "strict
homosexuality is more
<br>
> disordered than bisexuality." Reading it, however,
helped me remember
<br>
> that maladaption is part and parcel of adaption.
Disorder is part and
<br>
> parcel of order. The "mal" and "dis" prefixes are
nothing but
<br>
> value-laden subjectivity. The goo of reality extruded
through the mold
<br>
> of the author/thinker/subject. For someone like
Kirkegaard to claim
<br>
> they're being "objective" while using the "mal" prefix
is not even
<br>
> wrong. It's just bullshit. Apparently, my Woke Virus
infection is
<br>
> worse near my ears than near my eyes.
<br>
> </tangent>
<br>
>
<br>
> But the point is that *which* extended trait you use
(pencil, audio,
<br>
> text, etc.) chooses which interoceptive cycle you
engage. And when you
<br>
> pretend to make such a choice on purpose, at will, any
assignation of
<br>
> fault would be transitive. Which wolf do you feed?
<br>
>
<br>
> On 9/4/23 10:29, Steve Smith wrote:
<br>
>> I'm not sure my facility with the keyboard
actually serves me. As
<br>
>> many of you may suspect, and I suspect so myself,
it allows me to be
<br>
>> much less thoughtful and rigorous than I would be
in handwriting or
<br>
>> if I had some other throttle or impedance elements
between linguistic
<br>
>> centers and "paper"?
<br>
>
<br>
</blockquote>
<br>
</blockquote>
</body>
</html>