[COFF] Wikipedia anecdotes - LLM generalizations [was On the unreliability of LLM-based search results (was: Listing of early Unix source code from the Computer History Museum)

[COFF] Wikipedia anecdotes - LLM generalizations [was On the unreliability of LLM-based search results (was: Listing of early Unix source code from the Computer History Museum)

[Charles H Sauer (he/him)]

TUHS->COFF
>     > It's like Wikipedia.
> 
>     No, Wikipedia has (at least historically) human editors who
>     supposedly have some knowledge of reality and history.
> 
>     An LLM response is going to be a series of tokens predicted based on
>     probabilities from its training data. ...
> 
>     Assuming the sources it cites are real works, it seems fine as a
>     search engine, but the text that it outputs should absolutely not be
>     thought of as something arrived at by similar means as text produced
>     by supposedly knowledgeable and well-intentioned humans.
> 
> An LLM can weigh sources, but it has to be taught to do that.  A human 
> can weigh sources, but it has to be taught to do that.

Before LLMs, Wikipedia, World Wide Web, ... adages such as "Trust, but 
verify," and "Inspect what you expect," were appropriate, and still are.

Dabbling in editing and creating Wikipedia articles has enforced those 
notions. A few anecdotes here -- I could cite others.

1. I think my first experience was trying in 2008 to fix what is now at 
https://en.wikipedia.org/wiki/Vulcan_Gas_Company_(1967%E2%80%931970), 
because the article had so much erroneous content, and because I had 
worked/performed at that venue 1969-70. Much of what I did in 2008 was 
accepted without anyone else verifying. But others broke things/changed 
things, even renamed the original article and replaced it with an 
article about a newer club that adopted the name. A few years ago, I 
tried to make corrections, citing poster images at 
https://concerts.fandom.com/wiki/Vulcan_Gas_Company. Those changes were 
vetoed because fandom.com was considered unreliable. I copied the images 
from fandom to https://technologists.com/VGC/, and then citing those 
images was then accepted by the editors involved. (The article has been 
changed dramatically, still is seriously deficient, IMO, but I'm not 
interested in fixing.)

2. Last year, I created https://en.wikipedia.org/wiki/Hub_City_Movers, 
citing sources I considered reliable. Citations to images at discogs.com 
were vetoed as unreliable, based on analogous bias against that site. 
Partly to see what was possible, I engaged with editors, found citations 
they found acceptable, and ultimately produced a better article.

3. Later last year, I edited https://en.wikipedia.org/wiki/IBM_AIX to 
fix obviously erroneous discussion of AIX 1/2/3. Even though I used my 
own writings as references, the changes were accepted.

I still use the Web, Wikipedia, and even LLMs, but cautiously.

Charlie
-- 
voice: +1.512.784.7526       e-mail: sauer@technologists.com
fax: +1.512.346.5240         Web: https://technologists.com/sauer/
Facebook/Google/LinkedIn/mas.to: CharlesHSauer

[COFF] [TUHS] Wikipedia anecdotes - LLM generalizations [was On the unreliability of LLM-based search results

[steffen]

Adam Thornton wrote in
 <CAP2nic0sDT4K73n4Nc_29yt8h-fVgi+0u4NjErvobth6FUzUaw at mail.gmail.com>:
 |One could make a really good case right about now that:
 |
 |a) Vladimir Putin is doing his best to put the band back together, and

Sorry, sorry to come back on a computer list.  I stop thereafter.
But, one can only press thumbs that it all gets through without
further escalation; the desired block formation has resulted, at
quite a price, but the river is still floating.  If _there_
a really radical establishment would arise, though, and i mean
really really (really!), if it were only the military, you know,
there would be plenty of (pro) Russian(s) to protect in
Transnistria; they had to sit in the cold without heating this
winter, until a solution was found after months (and i *guess* and
*bet* it is a very expensive and other-side-favouring solution);
and their voted head is imprisoned, for, you know, in any case,
much much less than a jumbo.  You know; PJ Harvey sings

  On Battleship Hill's
  Caved in trenches
  A hateful feeling
  Still lingers
  Even now eighty years later

So all that can be said is.  Nothing changed, all the faults and
idiocies happen over and over again.  This is not "we have
a liftoff", but we are plain as stupid as we always have been.

(Btw.  In this house where i work we have a middle-age Ukrainian
Woman living here for many years, and a twin refugee couple.
Thirty meters aways there is a German/Ukrainian (married) couple
with lots of young children.  More Ukrainians, Russians and
Polish, all around here.  And i always said i admire the men, and
there are plenty, who deprive themselves from *that* government,
even if "their country is calling" them.  I only wish it would
worked out that way for me, personally and environmentally, would
i have lived during WWII.  And the opposite for WWI.)

 |b) Republicans are acting as extremely willing Russian assets, so therefore
 |c) Associating them with the red of Soviet Communism is pretty accurate.

The entire AI thing yet escaped me.  I never looked and tested AI.
I remember, on TUHS, and i think it was Steve Johnson who pretty
much enthusiastically talked about "thousands of 8-bit processors"
which understand a computer game "only by looking at the pixels"
so good that after several hours they "beat the milk out of" it.

[COFF] Re: [TUHS] Wikipedia anecdotes - LLM generalizations [was On the unreliability of LLM-based search results

[Alexander Schreiber]

On Sun, Jun 01, 2025 at 12:00:11AM +0200, Steffen Nurpmeso wrote:
> 
> There are surely useful tasks for AI, when it is driven with green
> energy, and after is has been fully understood.

What is currently being sold as "AI" is mostly LLM (Large Language
Models), which are - to grossly simplify things - massive brute-force
pattern matching engines.

There are plenty of use cases where a well setup pattern matching engine
is exactly what you need. My favourite example: SBB (Swiss Railways)
uses "AI" (an in-house trained pattern matching model) to sift through
the massive incoming stream of noise recordings from rail mounted
vibration sensors, to identify (by matching known qualified patterns)
those caused by damaged train carriage wheels. Additional support
infrastructure then identifies train, carriage, wheel and notifies
the owner/operator to fix the wheel - before gets worse and does
more damage to the rails.

There are lots of similar tasks where pattern matching engines are
a great fit (e.g. optical QC on finished surfaces during manufacturing).

If you try to use pattern matching engines for tasks that require
knowledge, thinking, understanding (in short: a trained human mind),
then you will be sorely disappointed while drowining in - potentially
even superficially plausible sounding - bullshit (See Harry Frankfurt,
"On Bullshit").

> Before that it is just another race that is raced at whatever cost
> there may be.  The price is payed by the environment, and the

The current forecasts of both the use and the costs (some claim that
we need to feed 90% of all power production to data centers eventually,
which is clearly .. ill advised) of "AI" are looking wildly exaggerated.

> grand children, but no longer further down the line.  That is
> possibly the good thing about it.  Exactly as said by the Club of
> Rome in 1972, and at least ever since also by the Catholic Church.

Club of Rome did not see a lot of scientific developments coming that
enabled growth way beyond their expectations. But then, predicting
the future is never a sure business.

That said, cheerfully burning down the planet for short-term profit
is not exactly a sustainable business model, to put it politely.

Kind regards,
           Alex.
-- 
"Opportunity is missed by most people because it is dressed in overalls and
 looks like work."                                      -- Thomas A. Edison

[COFF] [TUHS] Wikipedia anecdotes - LLM generalizations [was On the unreliability of LLM-based search results

[steffen]

Alexander Schreiber wrote in
 <aDxwjRhrUNJ5-Dm- at mordor.angband.thangorodrim.de>:
 |On Sun, Jun 01, 2025 at 12:00:11AM +0200, Steffen Nurpmeso wrote:
 |> There are surely useful tasks for AI, when it is driven with green
 |> energy, and after is has been fully understood.
 |
 |What is currently being sold as "AI" is mostly LLM (Large Language
 |Models), which are - to grossly simplify things - massive brute-force
 |pattern matching engines.
 |
 |There are plenty of use cases where a well setup pattern matching engine
 |is exactly what you need. My favourite example: SBB (Swiss Railways)
 |uses "AI" (an in-house trained pattern matching model) to sift through
 |the massive incoming stream of noise recordings from rail mounted
 |vibration sensors, to identify (by matching known qualified patterns)
 |those caused by damaged train carriage wheels. Additional support
 |infrastructure then identifies train, carriage, wheel and notifies
 |the owner/operator to fix the wheel - before gets worse and does
 |more damage to the rails.

That is possibly a great thing.  I can assure everybody that Swiss
freight trains pass by here in the many dozens / hundreds each
day, and they are very well maintained.  Most often they are
quieter than even the much light German passenger trains.
(Yes, there was that judgement in 2015 that the German (freight)
trains have to become silent by December 2020, before that it was
sheer unbelievable, especially during braking.  And i still can
recall the female DB speaker saying "und die Zeit brauchen wir
auch" "and that time we really need".  Unfortunately now the
brakes are a bit more silent, but wheel imbalance and wheel
bearings have massively increased (again).)
I say possibly because i could imagine sensors in the locomotive
should be capable to detect vibration irregularies?  Not that
i know.  But vibrations is understated given the hammerings.  Does
this really need sound recordings?  Interesting.

 |There are lots of similar tasks where pattern matching engines are
 |a great fit (e.g. optical QC on finished surfaces during manufacturing).
 |
 |If you try to use pattern matching engines for tasks that require
 |knowledge, thinking, understanding (in short: a trained human mind),
 |then you will be sorely disappointed while drowining in - potentially
 |even superficially plausible sounding - bullshit (See Harry Frankfurt,
 |"On Bullshit").
 |
 |> Before that it is just another race that is raced at whatever cost
 |> there may be.  The price is payed by the environment, and the
 |
 |The current forecasts of both the use and the costs (some claim that
 |we need to feed 90% of all power production to data centers eventually,
 |which is clearly .. ill advised) of "AI" are looking wildly exaggerated.

I also do not use cryptocurrencies btw.
I .. have heard Microsoft wants to buy a turned off nuclear plant
to reactivate it, only for training AI?  They also meddle within
the "small nuclear plants" scene.  I am thus against it.

 |> grand children, but no longer further down the line.  That is
 |> possibly the good thing about it.  Exactly as said by the Club of
 |> Rome in 1972, and at least ever since also by the Catholic Church.
 |
 |Club of Rome did not see a lot of scientific developments coming that
 |enabled growth way beyond their expectations. But then, predicting
 |the future is never a sure business.

I would think at the 50th anniversary of the publication there was
a profound review, and the pretty well got it.  That was the great
thing on Gore's campaign, in between the lines of that duel he
mentioned it.  With favouring "technology", of course, which
i never believed will outgrow the bad effects.  Not with that
overall mental state, anyway.

 |That said, cheerfully burning down the planet for short-term profit
 |is not exactly a sustainable business model, to put it politely.

Thank you.

 |Kind regards,
 |           Alex.
 |-- 
 |"Opportunity is missed by most people because it is dressed in overalls and
 | looks like work."                                      -- Thomas A. Edison
 --End of <aDxwjRhrUNJ5-Dm- at mordor.angband.thangorodrim.de>

--steffen
|
|Der Kragenbaer,                The moon bear,
|der holt sich munter           he cheerfully and one by one
|einen nach dem anderen runter  wa.ks himself off
|(By Robert Gernhardt)

[COFF] [TUHS] Wikipedia anecdotes - LLM generalizations [was On the unreliability of LLM-based search results

[als]

On Mon, Jun 02, 2025 at 03:14:30PM +0200, Steffen Nurpmeso wrote:
> Alexander Schreiber wrote in
>  <aDxwjRhrUNJ5-Dm- at mordor.angband.thangorodrim.de>:
>  |On Sun, Jun 01, 2025 at 12:00:11AM +0200, Steffen Nurpmeso wrote:
>  |> There are surely useful tasks for AI, when it is driven with green
>  |> energy, and after is has been fully understood.
>  |
>  |What is currently being sold as "AI" is mostly LLM (Large Language
>  |Models), which are - to grossly simplify things - massive brute-force
>  |pattern matching engines.
>  |
>  |There are plenty of use cases where a well setup pattern matching engine
>  |is exactly what you need. My favourite example: SBB (Swiss Railways)
>  |uses "AI" (an in-house trained pattern matching model) to sift through
>  |the massive incoming stream of noise recordings from rail mounted
>  |vibration sensors, to identify (by matching known qualified patterns)
>  |those caused by damaged train carriage wheels. Additional support
>  |infrastructure then identifies train, carriage, wheel and notifies
>  |the owner/operator to fix the wheel - before gets worse and does
>  |more damage to the rails.
> 
> That is possibly a great thing.  I can assure everybody that Swiss
> freight trains pass by here in the many dozens / hundreds each
> day, and they are very well maintained.  Most often they are
> quieter than even the much light German passenger trains.

Ah, but there is a difference: Swiss (SBB & Co.) passenger trains
are very quiet by design - because they carry passengers (aka
"cargo that can complain"). Swiss cargo trains are slowly getting
there as the rolling stock is replaced, but they were _not_ designed
to be quiet, as cargo tends not to complain.

> I say possibly because i could imagine sensors in the locomotive
> should be capable to detect vibration irregularies?

Won't work, because a sensor in the locomotive will have a hard time
recording noise from the end of the train that might be 100+m away.

The setup described about grabs a vibration recording of every wheel
as it passes the rail-mounted sensor and (with the help of other data
sources) the system then can identify the train/carriage/wheel.

>  Not that
> i know.  But vibrations is understated given the hammerings.  Does
> this really need sound recordings?  Interesting.

Well, it's vibration recording .. which is what sound is, just at possibly
a different frequency range that what humans hear.

Kind regards,
           Alex.
-- 
"Opportunity is missed by most people because it is dressed in overalls and
 looks like work."                                      -- Thomas A. Edison

[COFF] [TUHS] Wikipedia anecdotes - LLM generalizations [was On the unreliability of LLM-based search results

[steffen]

Alexander Schreiber wrote in
 <aEbbRayc6SoAJmaf at mordor.angband.thangorodrim.de>:
 |On Mon, Jun 02, 2025 at 03:14:30PM +0200, Steffen Nurpmeso wrote:
 |> Alexander Schreiber wrote in
 |>  <aDxwjRhrUNJ5-Dm- at mordor.angband.thangorodrim.de>:
 |>|On Sun, Jun 01, 2025 at 12:00:11AM +0200, Steffen Nurpmeso wrote:
 |>|> There are surely useful tasks for AI, when it is driven with green
 |>|> energy, and after is has been fully understood.
 ...
 |>|There are plenty of use cases where a well setup pattern matching engine
 |>|is exactly what you need. My favourite example: SBB (Swiss Railways)
 |>|uses "AI" (an in-house trained pattern matching model) to sift through
 |>|the massive incoming stream of noise recordings from rail mounted
 |>|vibration sensors, to identify (by matching known qualified patterns)
 |>|those caused by damaged train carriage wheels. Additional support
 |>|infrastructure then identifies train, carriage, wheel and notifies
 |>|the owner/operator to fix the wheel - before gets worse and does
 |>|more damage to the rails.
 |> 
 |> That is possibly a great thing.  I can assure everybody that Swiss
 |> freight trains pass by here in the many dozens / hundreds each
 |> day, and they are very well maintained.  Most often they are
 |> quieter than even the much light German passenger trains.
 |
 |Ah, but there is a difference: Swiss (SBB & Co.) passenger trains
 |are very quiet by design - because they carry passengers (aka
 |"cargo that can complain"). Swiss cargo trains are slowly getting
 |there as the rolling stock is replaced, but they were _not_ designed
 |to be quiet, as cargo tends not to complain.

Ah -- trains have an inside?  Interesting.
Hmmm, talking at least a decade.  SBB and also, i hate private
companies, BLS.  Mostly i guess chemicals via tank wagons, and
containers, lots of containers from an Austrian haulier.  I mean,
it is only snippets of time, when i am wandering in between, so
numbers are scaled at bit, but from impressions of all day and
night times.

Here pass by quite a lot, Swedish double-deckers, French TGV (they
renamed them, but slow german crawling..), German DB ICE of all
sorts, of course, but still normal ICs with mostly 2nd class, for
me it is Schnellzug and Eilzug but that cannot be helped,
wonderful Austrian trains with two red locomotives (great ones
which still aimed in record breaking performance, and deliver for
the quarter of a century), and in between them dove gray waggons
with red roof -- i love these trains, i think four times a day,
surely good cake and coffee within, their Nightjets.
BASF trains with Diesel and not FuelCell technology; we are back.

And lots and lots and lots of private freight train companies
which surely do not pay enough for being able to use the rails,
just for the sake of allowing "competition" .. at the cost of the
people, effectively.  And for ruining the name of the DB, which
actually pays for it.  Tax payers that is.

Loudest passenger train thinkable is by the way FlixTrain who
seems to run generators or what at one pair, but their bearings
and wheels scream to heaven "i have to live a long life" or what.
They surely spend not enough for using the rails, let alone for
the ear pain i have to suffer when that shit passes.  Whatever.
Luckily they pass by quick and do not brake, i cannot comment on
their brakes.

Swiss freight trains run smoothly and to a very large extend very
silent, and they break silent, too.  (Squealing happens at times,
for practically anything.  All the development for isolation, aka
decoupling, and materials technology aside.)

So, that is that.

 |> I say possibly because i could imagine sensors in the locomotive
 |> should be capable to detect vibration irregularies?
 |
 |Won't work, because a sensor in the locomotive will have a hard time
 |recording noise from the end of the train that might be 100+m away.

Mostly americans on this list i presume, and they still remember
the Wild West listen-at-the-rails trick, for whatever purpose.
Our rail bus with its diesel engine btw does not use engine
bearings.  "Boy the rail hums" quite a bit in advance.  So to say.

 |The setup described about grabs a vibration recording of every wheel
 |as it passes the rail-mounted sensor and (with the help of other data
 |sources) the system then can identify the train/carriage/wheel.

Well, .. sure.  I mean, in summertime i would consider talking the
job for an hour at weekdays, but i need tea, and i surely would
not be as exact.  For freight trains in particular, they are put
together at classification yards, often at little hills so that
they start rolling after some initial moment.  You know.
In Germany we say (or used to say, when we were still Germans)
"Where there is a will, there is a Way".  Something like that..

 |>  Not that
 |> i know.  But vibrations is understated given the hammerings.  Does
 |> this really need sound recordings?  Interesting.
 |
 |Well, it's vibration recording .. which is what sound is, just at possibly
 |a different frequency range that what humans hear.

It seems to me in rails it ranges further than in the air.

 |Kind regards,
 |           Alex.
 |-- 
 |"Opportunity is missed by most people because it is dressed in overalls and
 | looks like work."                                      -- Thomas A. Edison
 --End of <aEbbRayc6SoAJmaf at mordor.angband.thangorodrim.de>

--steffen
|
|Der Kragenbaer,                The moon bear,
|der holt sich munter           he cheerfully and one by one
|einen nach dem anderen runter  wa.ks himself off
|(By Robert Gernhardt)