[FRIAM] "I hope I'm wrong. But that text reads like it was generated by an LLM"

[steve smith]

I don't presume to guess whether others should or should not engage 
various types of technology to enhance their work (or play) product or 
maybe more critically, to enhance their own self-training.   It is 
perhaps never too late to have a good mentor and/or tutor on any skill 
or topic?

Keyboards (and soft terminals and printers and all forms of text 
processors, etc) have been my friend since about age 14 (typewriter) and 
while I *wish* had a more natural (or more trained) hand, my instinct is 
that I never had the right wiring for good penmanship, so the 
text-enhancing prosthetics have been a boon for me.  I was conversely a 
natural reader and naturally a good (but not perfect) speller and for 
the most part the mechanics of grammar came fairly natural (even if I 
didn't learn to diagram sentences as early as my Catholic School peers 
apparently did). Spell and even grammar checkers (esp. inline?) remain 
80% boon and 20% bane to me...  for those whose natural 
skills/inclinations are weaker, I am thankful they have that augmentation.

I second Glen's reaction, however ( if I understand it correctly) as 
refined or re-reflected by DaveW that when I detect an AI/LLM/ML 
generated layer over (or larded in?) to some text it undermines my 
confidence in who I am talking (listening) to and reading their 
"voice".   In the early days of Desktop Publishing, it was very 
disturbing to me to see lots of polished (looking) manuscripts which 
were at best first or second drafts... but I'm used to that now and can 
"read past it" and I *think* many people recognize the folly of using 
the polishing tools to polish their turds prematurely?   I think the 
same evolution will happen with generative AI products, probably much 
more quickly?

On a tangent brought up by Glen...  without searching back for the 
specific attribution/quote, I felt he took a swipe at the idea of using 
LLM's as "virtual friends" which in some manner I am very aligned 
with.   On the other hand, I am also defensive about it. I have referred 
to GPT as "my new bar friend" because that mode of conversation has been 
the one I most resonate with...   I only know "bar friend" because I 
have had them in the past, but my lifestyle/circumstance has taken me 
away from that milieu for years and I'm not likely to return to it. I 
envy Glen's "Salon" (at the local craft brewery?) but don't expect to 
re-establish or re-find that kind of experience anytime soon (ever 
again) myself.   As it is, GPT is quite a bit more adept at chatting 
"casually" with me on a much wider range of topics than the average Bar 
Friend.

My most recent "watering hole" was Tesuque Village Market, and that run 
lasted most of a decade, allowing me to establish rapport with 80% of 
the staff (and owners) and 50% of the regular clientele.   I knew who 
was available for what, and what to steer clear of with whom.   The old 
guy who could talk blacksmithing with me, but who never quite got over 
the idea that the only way to relate to a horse was through 
intimidation.  The British woman who designs women's shoes and travels 
the world from Milan to Brazil to get them constructed and distributed 
but still thinks Maggie Thatcher was as big of a hero as many today feel 
about Donald Trump (and perhaps for similar reasons)... or the suite of 
near-aimless 20/30-somethings trying to find their way in a world with a 
whole different set of opportunities and affordances than I experienced 
at their phase of life (the young and the reckless). The various 
professional couples whose conversational nature changed abruptly if one 
or the other was absent, exposing a whole layer of complexity in either 
not evident when they were in the homeostatic lock of the relationship 
(also interesting but different).  And on and on and on.

GPT chats with me about the wide range of design, material properties 
and sourcing, and construction strategies constrained by my desire to 
use as locally sourced, sustainable, low-tech materials as possible for 
the "old people's" addition I've been ideating for years.   I have one 
or two friends who can riff on some of this from time to time but they 
usually get hung up quickly on "but the standard/right/code-approved way 
is X!" and while GPT almost always starts out at the entrance to that 
dead-end of a rabbit hole, I can pretty easily persuade "him" to let go 
of that as a constraint and discuss the alternatives and the 
trade-space.   GPT is also quite willing and able to chat with me about 
a whole wide range of topics which I usually bring up here... and quite 
frankly is more openly responsive... GPT, for all it's hyper-utilization 
clearly doesn't have a "day job" so is willing and able to indulge me on 
just about any topic.

My biggest gripe about GPT (and ilk) is that it is too eager to please 
or agree with me.  I don't like it when it applies arbitrary limits 
(referencing DaveW's desire to get to GhostGPT) but I also am disturbed 
when it is too agreeable.  I really do want it to help play devils 
advocate when I conceive of various methods of insulating the ceiling in 
my addition with the mycelium of oyster mushrooms grown in variously 
loose straw, woodchips, recycled cardboard, hemp fibers.   I do want it 
to warn me of the potential issues (and mitigating strategies) around 
fireproofing, mold-prevention, ventilation, etc.      I have only one 
person in my life *interested* enough in discussing the traditional 
woodworking techniques for joinery (mortise-tenon, birdsmouth, saddle, 
etc) and *he* has no actual experience, just book learning and 
fascination.  GPT I suppose has only book learning too, but has read a 
*lot* more books and listened in on a LOT more conversations than my 
friend or I (I have practiced only a little post-beam joinery myself).

GPT chats with me at excruciating (for anyone but me or GPT) length on 
philosophical-unto-mathematical-unto-neurscience topics.   It has been 
willing/able to not only chat with me about any book I've read on a 
topic, it can usually do an excellent job of providing parallax across 
several authors/thinkers who may or may not have had formal engagement 
(e.g. referenced one another's work, etc.).   You might recognize some 
of my hobby horses in Terrence Deacon, Michael Levin, Karl Friston, Mark 
Solms, Jeff Hawkins, et al.

I suppose it is possible this is all just a huge computer-mediated 
circle-jerk... maybe that is precisely Glen's (implied) point?

PS...  on machine augmentation - I just received my Hypershell 
lower-body exoskeleton which was part of a Kickstarter I threw down on 
over 2 years ago.   I had hoped it might augment my hip-joint 
physicality enough to avoid surgical intervention but when it didn't 
come in on-schedule (6 mos from funding) I finally gave over and let the 
surgeon put me under the knife.  My second (right) bone-gristle hinge 
was cut away and replaced with a titanium/ceramic hinge in early 
December and 6 weeks later I'm still recovering.   The hip work healed 
as quickly as the first one (August) but the process aggravated an old 
(90s) "yoga injury", a rotated Lumbar joint which at the time yielded 
significant sciatic discomfort.    As the hip incision/bruising/etc 
recovered, it became evident that something more was afoot (pun 
recognized) as I had virtually no dorsiflexion and it was not recovering 
easily...  as the other pain receded I began to recognize that the 
residual discomfort/dysfunction was similar to what I had 30 years ago 
with the lumbar problem.   long story shortish, I'm now dragging one 
foot around as my brandy-new exoskeleton came in and I really cant 
properly test it out.  It is too lat to use to avoid hip-replacement but 
I still hoped it could provide enough augmentation/support for enhance 
my recovery... but the little use (test drive around the living room) 
suggests that while it can/might help with hip recovery, the 
dorsiflexion/foot-drop side effect undermines that.  Images of me with 
exoskeleton taking 6 strong strides then stumbling/tumbling only to get 
up and do it again... "forced error" if you will.  Not recommended by my 
surgeon.

Point to be noted... as a neo-retro Luddite, here I am using LLMs as bar 
friends and trying to cyborg myself back into mobility so I can hand 
build an addition out of raw vigas, fieldstone, cobb, and mushroom 
spawn, so I can live independently another 20 years so I can die of 
natural causes once all /natural causes/ have been outlawed by the 
AMA/FDA/NIH/WHO  WTF?  I guess I'll get back the thread on morbid ways 
to dispose of one's body once I've uploaded my psyche and soul to 
Altman's Stargate or Musk's GrokBoxx  or just AWS?

PPS  Extra nod to DaveW's reference to Stephenson's "Young Ladies 
Illustrated Primer"... I only wish I could conjure such for my 
granddaughter who really could use such a friend/mentor/protector as she 
struggles her way into/through the Tween Years in this modern era.  
Maybe some day if he lives long enough she can hang my rusty hypershell 
on the wall as a nostalgic artpiece the way I hang my grandfather's 
scythe and my parent's wood/leather snowshoes?

Don' let me get off on a (tangential) rant (rave?),

  - Steve


On 1/25/25 9:09 AM, Prof David West wrote:
> Pieter,
>
> I applaud your use of AI to improve your writing. It is my belief that 
> the*"proper"* use of AI, along with computers and computing tools in 
> general, is to _augment_ human abilities ala Vannevar Bush's "how we 
> may think," Douglas Englebart's institute, Alan Kay's dynabook, (the 
> fictional Young Ladies Primer of Stephenson's /Diamond Age/), and 
> Jobs' "bicycle for the mind."
>
> This is in direct contrast, it seems, to the sentiments of most on 
> this list who think that AI should, and inevitably will, replace 
> "inferior" human intelligence.
>
> I am curious if you see any question of "voice" in the AI improved 
> text? For example, I took glen's question as nothing more than an 
> observation that the "voice" of your post seemed to that of ChatGPT 
> instead of Pieter—definitely not *"augmented-Pieter,"* as I believe 
> you intended it to be.
>
> All of us respect Pieter and value his words. If, however, we are 
> confused by "voice." it raises issues of how much consideration the 
> writing should receive and how we should respond.
>
> Improving oneself, and one's writing, is a great goal. And we all have 
> experienced examples of "I wish I had said that," or "I wish I had 
> expressed that idea as eloquently as she did." If AI tools provide 
> wordings that you admire, or feel express your ideas more eloquently, 
> you should adopt them.
>
> Personally, I do not believe that AI can ever provide more 'eloquent' 
> writing, only more precise or more complete writing. So I, again, 
> strictly personally, would eschew using such tools as currently 
> constructed. I would however, if I had the chance, use Richard 
> Gabriel's tool, /Inkwell/, which, BTW, he emphatically states is not 
> an AI, to enhance my writing. But Richard is primarily a poet and 
> writer, despite his education and career in AI and computing, and he 
> created /Inkwell/ expressly to be a writer's assistant. I have 
> tremendous respect for Richard's writing and I know he uses Inkwell to 
> enhance his intrinsic abilities.
>
> davew
>
>
> On Fri, Jan 24, 2025, at 10:48 PM, Pieter Steenekamp wrote:
>> I was surprised by the comment "I hope I'm wrong. But that text reads 
>> like it was generated by an LLM"  At first, I just thought, 'so 
>> what?' But it got me thinking about how AI changes how we 
>> communicate, which is really important to me.
>>
>> Here's my main point:
>>
>> I think it's okay to use others, whether they're people or AI, to 
>> help me communicate better, as long as the ideas are mine and I'm not 
>> copying someone else's work. If using AI to polish my words bothers 
>> someone, we can talk about it.
>>
>> Here's some background on why I think this way:
>>
>> a) Learning to communicate is one of the best things you can take 
>> away from school or university. For example if a student is supposed 
>> to use her own words in an assignment, then using AI or asking 
>> someone else to rewrite is wrong.
>>
>> b) In the workplace, we've always had people like personal assistants 
>> or speech writers. I see using AI in the same way, just a modern tool 
>> to help.
>>
>> c) Writing assistance used to be expensive because you needed people. 
>> Now, AI can do the job for free, and that doesn't go against what I 
>> believe is right.
>>
>> d) Your writing should still sound like you. So, I write something 
>> first, then ask AI to make it better but keep it simple and in my 
>> style, like, 'Please make this sound better but keep it easy to 
>> understand.' I now realise I violated this previously and will adapt.
>>
>> e) Maybe it's good, depending on the situation, to mention when 
>> you've used AI in your writing.
>>
>> f) Here's a funny story: I used AI to help write a message for an AI 
>> workshop, and someone complimented the writing so much they suggested 
>> I should be a writer instead!
>>
>> My bottom line:
>>
>> I plan to add a note at the end of each message in this group saying 
>> I used AI to help with my writing. If my using AI is a problem for 
>> anyone here, feel free to ask me to leave the group or simply remove 
>> me from the group.
>>
>> Note: I use AI to assist my writing.
>> .- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. 
>> --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-..
>> 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/
>>
>
>
> .- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-..
> FRIAM Applied Complexity Group listserv
> Fridays 9a-12p Friday St. Johns Cafe   /   Thursdays 9a-12p Zoomhttps://bit.ly/virtualfriam
> to (un)subscribehttp://redfish.com/mailman/listinfo/friam_redfish.com
> FRIAM-COMIChttp://friam-comic.blogspot.com/
> archives:  5/2017 thru presenthttps://redfish.com/pipermail/friam_redfish.com/
>    1/2003 thru 6/2021http://friam.383.s1.nabble.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20250126/d062926d/attachment.html>