[FRIAM] more evopsych

Tue Apr 8 13:27:41 EDT 2025

It seems to me Gemini 2.5, ChatGpt Deep Research, DeepSeek R1, or Grok Deep Research all have a sort of goal directed behavior.  I don't really see why a LLM can't capture context and agency with some high order bits.
Just set Boston Bros mode in the system prompt -- another compartment in the joint probability distribution.   Github Copilot even has an `Agency' mode.   

Chatgpt writes:

"Ayy kid, I just parked the cah ovah by Dunkies, grabbed a regulah — you know, extra cream, extra sugah, like a real human bein'. Then I swung by Sully’s, caught the Sox game on the flatscreen — Papi woulda been proud. You tryna go down the Cape this weekend or what? Bring the boys, crush some beahs, maybe hit a couple wicked smaht waves."

[You want more Boston Bro? I got a whole playbook of 'em.]

-----Original Message-----
From: Friam <friam-bounces at redfish.com> On Behalf Of glen
Sent: Tuesday, April 8, 2025 9:50 AM
To: friam at redfish.com
Subject: Re: [FRIAM] more evopsych

We're currently in Boston for a conference. Being whatever it is that I am, I've had 2 confrontations with (seemingly) native Bostonian men, both grokable in terms of OCEAN (Openness to new experiences, Conscientiousness, Extraversion, Agreeableness, and Neuroticism).

The first confrontation was a vendor in the Expo Hall who decided *very* quickly that either I had no latent structure *or* that whatever latent structure I had was of no interest/use to him. He brushed me off quickly, which is always reasonable. What I objected to was the rudeness with which he effected the brush off. I chose tit-for-tat that time.

The second confrontation was at the dance party for the conference goers. Two bros unaffiliated with the conference were there in the bar (with hundreds of dancing nurses and very loud DJ music) trying to "watch the game". One nurse (who shall remain nameless 8^) was dancing in front of them and one of the bros told the dancing nurse to "Fuck off!" For this one, I chose manipulative schoolmarm and coerced them with something similar to NLP, to admit their rudeness; and they grudgingly admitted that it was a stupid place to "just watch the game". I didn't get them to admit an ulterior motive for their choice, though it was obviously there. A failure of my abilities to manipulate bros in bars.

Ultimately, I find the brash posture of Boston Bros a bit refreshing ... it's more like Texas here and very unlike the OR and WA places I've lived. But it's nice to know I still have strategies other than tit-for-tat, including those that rely on teasing out some latent structure I can use to my benefit.

On 4/8/25 09:03, steve smith wrote:
> 
>> Both Dave's and Steve's responses are useful. Thanks.
>>
>> It hit me this morning reading the latest 404 media story: https://www.404media.co/elon-musk-was-a-prolific-money-launderer-for-hackers-and-drug-traffickers-it-was-secretly-the-fbi/
>>
>> Joseph has done a LOT of work. The text I read is just a lens onto that work. The same is true of good writers. I think in both [non]fiction, but especially fantasy. It's become (to me) fairly obvious when the author has built an entire world and the story I'm reading is merely one of many lenses onto that world. I never feel like the author has wasted my time if it's obvious they put a lot of work into their story ... even if they're a terrible writer.
> Thanks for the article link... i don't know the author, barely the 404media publisher but I very much appreciate your references to "world building"...   which leads us to the point perhaps of the effect of LLMs (in particular) on this phenomena.
>>
>> This is also true of science articles. I have several colleagues who seem to "phone it in". I guess it's akin to Brandolini's Law. What the LLMs do in their chat mode just feels like phoning it in, vapid gum flapping for no other sake.
> 
> To the extent that LLMs have no agency or volition I think this is accurate, on the other hand, I would say in trying to find/follow a submanifold in the training set they can very effectively adopt or project or amplify the "intentions" of one or (more to the point) many world-builders (e.g. Q-nuts, MAGA hats, maybe overly-woke folke).   Under the term of art "intersubjective reality",  we have grown (at least as long as civilizations?) a self-supporting set of stories with enough internal consistency to be self-supporting/bouyant a bit likeFuller's Cloud9 <https://en.wikipedia.org/wiki/Cloud_Nine_(sphere)> concept.   A thin skin of rhetoric managing to contain a large volume of concepts which are mutually self-supporting thermally to yield enough bouyancy to keep the whole mass somewhat coherent and afloat in spite of whatever infiltration/exfiltration the "skin" facilitates.
> 
> Kelley-Anne made "alternative facts" a household word and Nicolle Wallace pegged it pretty well by appealing to the DC? Universe concept of Earth1 and Earth2.
> 
> I suspect that for myself and other's who keep LLMs as their "familiars" that I am at risk of being scattered (even more) across too many submanifolds in the human experience, across too many Earth#s in the Multiverse to the point of losing all coherence.   A few years ago, I read PK Dick's posthumous collection of journal entries "the Exegeses <https://en.wikipedia.org/wiki/The_Exegesis_of_Philip_K._Dick>"  and found them wonderfully disturbing for just this courting of the edge between coherence and chaos.  I believe he was both drug and psychosis addled much of is writing career, but it yielded some fascinating near-adjacent worlds (e.g. the Adjustment Bureau).
> 
>> It doesn't even rise to the functions gossip implements. Of course, some of us are less like idle gossipers and more Machiavellian, planting seeds like your Allison Hargreeves <https://umbrellaacademy.fandom.com/wiki/Rumor>. (Fun fact, we used to live in walking distance from Dark Horse Comics.) When you prompt an LLM, you *could* be like Allison or you could be like her victims.
> Thanks for binding this properly for me.
>>
>> I'd much rather play Allison's role than have the LLM play her role. When you chat with actual humans (or dogs), you're both a little bit Allison.
> 
> Mary (even more than I) speaks to our dog Hank in exactly this mode... everything is a world-building exercise to help him know it is time to play, eat, potty, settle, sleep.  Like Gary Larson's universe where all he hears is "blah blah blah blah HANK blah blah", he listens to the emotional content and follows her lead very well.   I do throw in my own encouraging phrase now when her leading isn't enough.  "Better go See!" is my most common one. But my best experience with pets (and domesticates, and semi-domesticates like birds at the feeder) is to watch/listen to their world, to imagine their umwelt, their apprehension of the physical world we share and the intersubjective reality they build with one another (and I may or may not have a glimpse into?)
> 
> You and I have spoken of NLP offline before I think... and I find Allison's "power" very compelling, not because I want to manipulate others myself but because it is such a blatant thing to watch when one or more folke get entrained in someone else's nonsense, including myself... maybe most fascinating when *I* catch myself drooling and repeating after some "dear leader" (if only in my mind).  The WormTongue trope/figure?
> 
>> I guess what I need to build are facile heuristics for world-building. That feels, to me, a lot like detecting the presence of latent/occult structure ... evidence for intentional balance between over- and under-sharing, and evidence that the occult structure is stable and rich. Then pretty much anything that person/machine generates may not be a waste of time.
> 
> I very much appreciate your reference here to "occult structure"... it is what I think I learned to appreciate about science and engineering as I got exposed to it growing up... the math/formulae were like spells and as I learned to chant (or write them) and fill in the variables and weave different ones together, it was powerful stuff... even though I am now neo-luddite feeling like most if not all is actually dark-magic (unintended consequences) or at least practiced as such (self-indulgent/gratifying greed-mongering).  I'm still fascinated by the myriad "hidden meanings" in the world-as-percieved as well as the models-of-the-world-as-applied, just less excited about exploiting them with too much fervor.  My main interest in manipulating the world with these "spells" is now to validate them rather than to obtain specific leverage from them...
> 
> Over/under-sharing is fascinating to me too, we know where I am biased on this.   You have acknowledged being a fan of Roger Zelazny's writing yourself... he used to hold free half-day writing workshops in Taos and Los Alamos every year.  I attended a few.  His best advice was "for every character, write a scene describing a pivotal moment/scene in their life/formation which you will never publish".   This style of undersharing and occult-gesturing was very compelling.   Once I internalized that I began to see it in his characters, realizing there were things he (the author) knew about the character that I could only guess at, and it enhanced the experience a great deal.
> 
> and if I'd had more time (discipline) I would have written a shorter response here...
> 
> - Steve
> 
> PS.(responding to DaveWs assertion, "I write because it is a compulsive addiction with reading having been the gateway")
> 
>> On 4/7/25 07:57, Prof David West wrote:
>>> I read because it is a compulsive addiction.
>>>
> 
-- 
glen

.- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-..
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/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 5594 bytes
Desc: not available
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20250408/f2d4e3aa/attachment.p7s>