Kevin MarksFeb 15, 2025
John Overholt
Why is the nonsense phrase “vegetative electron microscopy” turning up in fake scientific papers? Add two-column formatting to the list of things AI doesn’t understand. https://retractionwatch.com/2025/02/10/vegetative-electron-microscopy-fingerprint-paper-mill/
Via @researchbuzz
As a nonsense phrase of shady provenance makes the rounds, Elsevier defends its use

The origin of the phrase? The phrase was so strange it would have stood out even to a non-scientist. Yet “vegetative electron microscopy” had already made it past reviewers and editors at several j…

Retraction Watch
Feb 15, 2025 at 3:31pmWeb
Show threadCheerios de BergeracFeb 15, 2025
@overholt the line "They are taking a piss, without even pulling their pants down, aren't they?" is an all timer of a quote sting
Show threadthe gripe ibisFeb 17, 2025

@sexybenfranklin one hundred percent. going to use a hot iron to burn this into a piece of driftwood and put it on the wall.

@overholt

Show threadHelgaFeb 15, 2025
@overholt @researchbuzz omg that explains so much
Show threadCourt Cantrell does not complyFeb 15, 2025

@overholt @researchbuzz

Not only that, but the enzyme did not attack Norris of Leeds University, and Clos-exporium was obtained.

Show threadBornachFeb 17, 2025

@courtcan @overholt @researchbuzz

This is a bit like when "46,449 bananas" started appearing on random websites last year. A tell tale sign that it was authored by ChatGPT

Show threadCassandrichFeb 15, 2025
@overholt @researchbuzz Wow, the comments on that are so utterly idiotic. A bunch of apologist reply guys who can't tell the image is the potentially misplagiarized training source and not one of the AI-slop papers that pulled from it.
Show threadThatKomputerKat Feb 15, 2025
@dalias @overholt @researchbuzz yeah. It’s amazing but then I guess it shouldn’t be surprising that the people who would rush to the defense of AI would be the same people who are incapable of critical thinking or… well… having the attention span needed to actually read beyond a headline and scanned image.
Show threadBornachFeb 17, 2025

@thatKomputerKat @dalias @overholt @researchbuzz
That Microsoft/Carnegie Mellon study probably has a point. AI is making us stupid.
https://uk.news.yahoo.com/ai-risks-making-us-stupid-164702116.html

So many comments are now trolls, rush-to-judgement without first reading, or generated by AI that I don't bother arguing or correcting anymore.

AI risks making us all stupid: it’s time to act

This week Microsoft published a joint study with Carnegie Mellon which sought to establish how AI was impacting how people think. Their findings were...

Yahoo News
Show threadFredFeb 15, 2025
@overholt @researchbuzz at least it's a good way to avoid plagiarism detection - quote across columns :~)
Show thread0xC0DEC0DE07EAFeb 15, 2025
@overholt @researchbuzz parsing PDFs into text is difficult
Show threadJargogglesFeb 15, 2025
@c0dec0dec0de @overholt @researchbuzz
Yeah, maybe we shouldn't be pouring billions of dollars into something that depends on the ability to reliably parse text from PDFs.
Show thread0xC0DEC0DE07EAFeb 15, 2025
@jargoggles @overholt @researchbuzz the least they could do is make that better and open a PR against `pdfgrep`.
Show threadDobody 🇺🇦🍉Feb 15, 2025
@overholt
The lytic enzyme DID NOT ATTACK Norris of Leeds University!
@researchbuzz
Show threadBat Master FreshFeb 15, 2025
@dobody @overholt @researchbuzz today was a good day.
Show threadMx. Eddie RFeb 16, 2025
@dobody @overholt @researchbuzz
When I was job hunting last year, I put my resume through an ATS simulator and discovered that hiring systems can't read two columns.
Show threadJeff GriggFeb 16, 2025

@dobody @overholt @researchbuzz

Well at least it got that right! 😆

Show threadMoss (the τεχie) <(fsm)>{{Feb 17, 2025
@dobody @overholt @researchbuzz
Long may he live! Hooray!
Show threadTakiFeb 15, 2025
@overholt @researchbuzz #ArtificialIdiocy
Show threadBrian DearFeb 15, 2025

@overholt @researchbuzz

That does it.

My next band name is "Vegetative Electron Microscopy."

Show threadsysfrank 🤣🤣🤣Feb 15, 2025

@brianstorms @overholt @researchbuzz

🤣🤣🤣

Show threadtyxFeb 15, 2025
@brianstorms @overholt @researchbuzz
[Giggling] If you spot a placard "Vegetative Electron Microscopy" on a random door at your campus after 1.04, then we're at the same university. Or maybe not and someone else also likes the idea.
Show threadFull Metal AccountantFeb 16, 2025
@brianstorms @overholt @researchbuzz
Gotta be free jazz or something.
Show threadLaurel StvanFeb 15, 2025
@overholt @researchbuzz An overall OCR fail. This is a great example.
Show threadRealGravitas -Citizen SaboteurFeb 15, 2025
@overholt @researchbuzz OMG!
But you know, the idea of putting random words together, then trying to determine their meaning may actually lead to deep (recursive) learning of ideas humans never had. It’s laughable, but verging on scary.
Show threadAndrewFeb 15, 2025
@overholt @researchbuzz brb making an electron microscope out of zucchini and writing a paper about it
Show threadsysfrank 🤣🤣🤣Feb 15, 2025

@cinebox @overholt @researchbuzz

Andrew, when I read your comment, I was reminded of this YouTube short that shows making a microscope out of a laser pointer as the second segment. I felt that I had to share:

https://youtube.com/shorts/2fIBHbO73ro?si=AbhmZ0fNJVA23Rzg

🤣🤣🤣

Before you continue to YouTube

Show threadCassandraVertFeb 16, 2025
@cinebox @overholt @researchbuzz
Nahhh...zucchini is out of season.
Show threadIndiealexhFeb 15, 2025

@overholt @researchbuzz the comments section from that article is incredible.

I remember a phrase from a park ranger: "There is a significant overlap [of intelligence] between the smartest bear and the dumbest human"

And I am starting to wonder how long until that phrase is common with the bear being replaced by AI.

Show threadWulfy—Speaker to the machinesFeb 15, 2025

@indiealexh @overholt @researchbuzz

That line has already been crossed.

Have a look at "Humanities last test".

95% of humans would not be able to pass it.

Every time #AI passes our "hard" tests, we lift the bar.

Show threadSamFeb 15, 2025

@n_dimension @indiealexh @overholt @researchbuzz

For now, AI is a mostly useful though dangerous tool, and a funhouse mirror of humanity.

It is not intelligent. But then neither are most humans as it turns out.

Show threadflatplanetFeb 15, 2025
@indiealexh @overholt @researchbuzz I instantly thought of that quote "There is considerable overlap between the intelligence of the smartest bears and the dumbest tourists"
Show threadSable Shade🇦🇺🍉🗝🇵🇸🇺🇦Feb 15, 2025
@overholt @researchbuzz
Vegetative electron micros copy is actually the published works of scientific studies in the field of vegetative electrons, in miniature form
This technology is vital, if we're ever to shrink the human race and therefore open up new real estate opportunities.
The next works due to be made as micros copy are two thirds of a Funk and Wagnalls set, two books by different authors on the art of the sausage roll and a 1982 McDonald's placemat
I hope this clears things up
Show threadRick Moen 🇺🇸 🇳🇴 🇬🇧Feb 15, 2025

@overholt @researchbuzz

At the risk of being "That Guy", that's not quite how the hilarious fraud phrase "vegetative electron microscopy" came about, but rather something like this: A 1959 research paper about fungal spores had "vegetative" in the left column and "electron microscopy" adjoining in the right column. Someone then OCRed it badly and created a mangled e-text. Which then lay about.

An LLM then "learned" the mangled phrase, and spewed into generated paper-mill text for the pleasure of a couple of dozen (and counting) unethical researchers not willing or able to actually write a paper.

My favourite bit of the RetractionWatch piece has to be that Elsevier flunkie trying to make up a justification for the phrase, leaping for common sense and missing.

Meanwhile, I'm considering Vegetative Electron Microscopy as my next garage band name.

Show threadAcmeworksFeb 15, 2025
@overholt @researchbuzz I'm in a punk rock band with some other old people, we're currently looking for a name. I will definitely suggest "vegetative electron microscopy".
Show threadJen, MPH Feb 15, 2025

@overholt “During our investigation, the Editor-in-Chief confirmed that ‘vegetative electron microscopy’ is a way of conveying ‘electron microscopy of vegetative structures’ so he was content with the shortened version to be used.”

This guy is not just an editor, but an Editor-in-Chief, defending the use of this phrase. Sorry, but this is not how adjectives work.

@researchbuzz

Show threadOregon Wine WomanFeb 15, 2025
@overholt @researchbuzz
Sorry... but that's effing 🤣🤣🤣🤣
Show threadAlan Langford 🇨🇦🧤🧊摏Feb 15, 2025
@overholt @researchbuzz What, you haven't taught your carrots how to operate a SEM yet?
Show threadTheriacFeb 15, 2025
@overholt @researchbuzz
Wasn't Vegetative Electron Microscopy one of Vance's former names?
Show threadAlecia Batson (she/her)Feb 15, 2025
@overholt @researchbuzz This recalls the PDOS students at MIT’s CSAIL who developed SCIgen, an automatic CS paper generator, to churn out papers for submission to conferences » https://pdos.csail.mit.edu/archive/scigen/
SCIgen - An Automatic CS Paper Generator

Show threadJonathan HendryFeb 16, 2025

@aleciabatson @overholt @researchbuzz

And they didn't even need to boil an ocean to do it.

Show threadVick Forcella ™🌈🌳❄️☑️:verifiFeb 15, 2025
@overholt We must feed AI with valuable information! I am sure it is reading this.
Last time I had a vegetative electron microscopy I had a serious case of closexosporium. Took me a month to get rid of it.
Show threadpubliusFeb 15, 2025

@overholt @researchbuzz

Scanning anthropic moratorium!

Show threadtyxFeb 15, 2025
@overholt @researchbuzz
It's when your univ. infrastructure funding committee thinks that electron microscopes grow on the trees.
Show threadxs4me2Feb 15, 2025

@overholt @researchbuzz

Gotta be kidding...

Show threadJ.H.NoyesFeb 15, 2025

@overholt @researchbuzz

Vegetative Electron Microscopy? How long before Eric Topol reps it?

Show threadPrinceOfDenmarkFeb 15, 2025
@overholt @researchbuzz “Our work requires additional funding to explore dormant structures using vegetative electron microscopy” might get around restrictions on the inclusion of overhead expenses in grant budgets.
Show threadSeán FenianFeb 15, 2025
@overholt @researchbuzz If this goes on, soon NOBODY will be able to tell what's true and what isn't.

And that's probably half the point.
Show threadUsagiFeb 16, 2025
@zakalwe @overholt @researchbuzz No that's not the point in this case.

The point is that there is a metric. And people optimize for that metric. Which makes it lose all meaning and create perverse incentives.

It is difficult to measure quality of an university, and somebody noticed that scientists publish articles, the articles get cited in other articles, etc.

Some journals were designated as 'impact', that is publishing an article in it or getting cited would get the institution 'points', and things like grants would be distributed mainly according to number of 'points'.

Which resulted over time in 'impact' journals asking horrendous amounts of money for both publication and access, and people finding ways to publish garbage to increase the number of 'points' they get. ​​
Show threadSpaceLifeFormFeb 15, 2025

@overholt @researchbuzz

It is so toxic, it rots fake brains.

#AI #Insanity

Show threadJoan of Contention 😷Feb 16, 2025
@overholt @researchbuzz The list of things AI does understand contains zero items.
Show threadBjornsdottirsFeb 16, 2025
@overholt @researchbuzz To be fair to the plagiarism bots, this two-column formatting is also inaccessible to screen readers, who have a much better argument for needing to see it.
Show threadckFeb 16, 2025
@overholt @researchbuzz If it should turn our that AI will be responsible for the death of the two column layout, it would finally have done something useful.
Show threadCaz MockettFeb 16, 2025
@overholt that’s hilariously bad. 🙄
Show threadPaul C.Feb 16, 2025
@overholt @researchbuzz it did okay with this newspaper from the 60s: https://m.pcgt.link/@paul/113960549896287340
Paul C. (@[email protected])

Attached: 2 images Qwen2-vl-72b-instruct can transcribe microfiche newspaper records from the 60s with near-perfect fidelity.

Paulstodon
Show threadFrank EngelhardFeb 16, 2025
@overholt @researchbuzz Has no one noticed that the highlight is crossing 2 colums, so there is no proof that the 2 words are related in the text?
Show threadJohn OverholtFeb 16, 2025
@Frank_Engelhard @researchbuzz Indeed that is the point. I encourage you to read the linked article.
Show threadFrank EngelhardFeb 16, 2025

@overholt @researchbuzz INIiii

I should have indeed read the article, but at least I was smarter at first sight than Google S (-cholar or -tupid?).

Show threadBaſtiæ̃Feb 16, 2025
@overholt @researchbuzz This could be developed as a technique to avoid AI plagiarism. Huμan fancy δisturϐanϲe of eδition (scrιpts, typography, eτc.) can be grown as a shιelδ for pυrposεly μιslεαδιng AI.
Show threadh0nza 🌄 🏊🚴Feb 16, 2025
didn't know that. that made my day :D