ubi
TattieLLMs are mansplaining as a service, but more specifically that type of mansplainer who googles your question and replies authoritatively with the first result that comes up, despite having zero understanding himself.
At least with a Google search you can have a squint and see if the information comes from some government department or "homeschool-moms.com". LLMs flatten both the source and the intent of the information into the same officious tone, which is why you see them echoing Reddit shitposts as fact.
"All the information is on the internet"
Not all information has the same value, motherfucker.
True expertise is in learning to distinguish and contextualise. And preserving that meta-information is a vastly undervalued skill.
(Salutes librarians)
Not all information has the same value, motherfucker.
True expertise is in learning to distinguish and contextualise. And preserving that meta-information is a vastly undervalued skill.
(Salutes librarians)
Anyway this is why the integrity of human knowledge against the malignant growth of LLMs will be fought for, and won, by shitposters.
Kat@Tattie You just reminded me of a movie-special-effects guy I met a long time ago.
He had a pitch for an educational TV show. The idea was that a group of schoolchildren would be brought in, and a scientist(-looking) person would explain something to them.
The twist was that only about the first half of the presentation would be strictly factual. They'd start by including subtle errors, move on to less-subtle factual errors, building up to completely OTTWTFBBQ fantasy ravings by the end.
The thrust of the setup was to see how long it would take the kids to twig that something ain't right, and to then speak up and start questioning it. The aim, of course, was to teach them (and the audience) to listen critically and question anything that smelled off, no matter how well-credentialled the speaker might be.
Sadly (and predictably) it never got optioned, let alone actually implemented.
I feel like we could really use that kind of thing... about 20 years ago, actually. Or maybe even sooner.
@Tattie Irrelevant but amusing aside: his company made TV adverts for Bundaberg rum, featuring a life-sized polar bear man-in-suit. You know, just the sort of thing you'd expect to find in tropical Queensland.
Said bear was idolised by many chest-beaty manly-men as the epitome of grunt-snort manliness.
The team were endlessly amused by the thought of how these guys would react if they found out that the man in this suit in fact a)was very, very gay, and b)had the physical appearance of about five and a half feet of knotted string.
eswillwalker
Ozzelot
notsoloud
Reiner Jung@Tattie exactly.
I came across a text from 2023 recently that used that term. And I think it is the best term to describe LLMs in their current form as they are pushed I to every product.
But here is the source
https://ryanleetaylor.com/blog/mansplaining-as-a-service
miniBill (Leonardo)@Tattie so what you're saying is that we should Shitpost Harder?
Oscoder
Victor Gijsbers@Tattie "Where is the knowledge we have lost in information?"
Benoît Régent-Kloeckner@Tattie All the information is also (probably) in the decimals of pi, for that matter.
@BrKloeckner including the schematics for building that AI god that the techbros are seeking... if you know which digit to start from 😂
UnlikelyLass@BrKloeckner @Tattie I remember when the DVD DRM was cracked and someone worked out where in Pi it was, as a kind of 'steganography' for communicating it legally.
It made me think that we could encode anything that way -- binaries, images, etc. Of course, you do have to *find* it in there...
Nick
Jessamyn
Mx Verda
Dan W
Steve Williams
noplasticshower
feld
Mads (DBD 🏳️⚧️)@Tattie That's an absolutely fantastic way of describing it
Stephen Elliott~Buckley
PhosphenesSpamsplaining.
Patrick H. Lauke@Tattie hah, was thinking along the same lines (for the second part) earlier https://mastodon.social/@patrick_h_lauke/114805904694167439
mk30@Tattie @ordinoides and that's why I call him ChadGPT.
thurisaz@Tattie "Mansplaining as a service" is one of the best descriptions of LLMs I've heard. Thank you!
Paul Plowman@Tattie This is absolutely the best way of describing LLMs I've heard.
Chris J. Line@Tattie The Ultracrepidarian Machine
Chance 'Um@Tattie I dislike the term “mansplaining” but love the analogy.
Daniel Bohrer@Tattie now there's a worse type of mansplainer: the type who puts your question into ChatGPT and replies authoritatively with the first answer that he gets without questioning it or having zero understanding himself.
@daniel_bohrer it's mansplainers all the way down!
rowan {Sierra sunshine woman}@daniel_bohrer I work with a few of those people. They'll plop an AI answer in to the chat (sometimes that's only vaguely tangentially related to the actual question), and the number of times they'll come back a few minutes later with "ooops, it lied" is astounding.
stibbons@daniel_bohrer @Tattie unfortunately I've seen folk say "let me chatgpt that for you" unironically.
Geoff 🏴@Tattie I've got a worse iteration. A job-based forum where a particular user regularly posts the LLM reply to "I've got a question about XYZ", with "I hope this helps". A lot of it is safety critical stuff, and yes, there's errors. I'm waiting for the first instance of a loop where Google's LLM supplies the text generated by ChatGPT on that forum as the first result.
an actual bus@Tattie Also the guy who confidently asserts that your thing "won't work" even though it... does.
Chuck
ajuvo ✔
Joanna Holman@Tattie it is taking all my restraint not to tell the people who mansplain by pasting what chatGPT told them into an online conversation and think they’re being helpful what I think of them
Tim Clevenger 
@Tattie "Large Language Mansplainer"
Evan
Rowan Brad Quni
Pēteris Krišjānis@Tattie zing 👌 perfect explanation.
vekkq@Tattie yet, you asked it, probably already knowing where the answer would go and without defining a better tone, despite being annoyed by it.
Michael Harley
HGourlayUCL@Tattie Yes. An annoying interrupter that jumps in with 'I can summarise this article for you'. (A) Not usually the type of doc that would benefit from summarising. (B) Totally unwanted as what I wanted was to quietly read and make meaning from the article without interruption.
@HGourlayUCL ugh, when Adobe Reader started popping that up, I was so damn annoyed! 🤬