Sarah L. Fossheim

A while ago I was talking to a researcher and he was like “a few years ago through field research I found out that __ and wrote about it, it was such an exciting find” and then added “and it’s amazing how smart ChatGPT is, recently I asked it the same thing, and it came to the same conclusion as I did in my research!”

My man, ChatGPT didn’t come to the same conclusion you did and it isn’t smart. It stole words from your article and didn’t give you credit.

That’s not amazing, that’s plagiarism.

Feb 18, 2025 at 9:47amWeb
Show threadSarah L. FossheimFeb 18, 2025
the world is doomed 
Show threadEndlessMasonFeb 18, 2025
@fossheim
You think if we all post "your chat gpt licence has expired" the bots will start telling people that?
Show threadCorvusFeb 18, 2025
@EndlessMason @fossheim Character encoding allows to do fancy stuff. Like text that is there but not visibly rendered. So, maybe that is how this can pulled off. Machines parse the hidden message, human moderators do not see them. Sadly I do not recall the link the the article explaining how this is done.
Show threadEndlessMasonFeb 18, 2025

@corvus_ch
Stenography?

@fossheim

Show threadFiona BianchiFeb 18, 2025
@fossheim like a lot of technology, many people don’t understand it in the slightest. I think things are often perceived as miraculous because limited exposure often shows only the good points. Most engineering endeavours, particularly extremely complex ones, have quite marked limitations due to basic laws of physics, chemistry, mechanics etc. It’s like the swan 🦢 metaphor - you see a graceful thing gliding over the water but if you look underneath there’s legs frantically paddling. Same with a lot of IT: what you see on the screen is amazing but to get that you have a lot of simple, repetitive, code being run at frantic speeds. AI is the epitome of this accept that the above-water “grace” requires thousands of swan legs. It’s not efficient and it’s not pretty. Humans on the other hand, can be vastly more efficient and get better results.
Show threadMeercat ✅ 
Feb 18, 2025
@fossheim what I wonder in these cases is that why it is not obvious to the people. Of course it is plagiarism
Show threadAndrocatFeb 18, 2025
@fossheim It's incredible how someone can be in science and not understand the first little bit about source critique...
Show threadggdupontFeb 18, 2025
@fossheim yet another case of people seeing intelligence where there is none, just access to lots of data... a kind of pareidolia for intelligence
Show threadmo7emmedFeb 18, 2025
@gdupont @fossheim
AI IS intelligent. Just come up with a new logical riddle, and you will find that in most cases it can solve it.
Show threadggdupontFeb 18, 2025

@mo7emmed @fossheim yet another case of people seeing intelligence where there is none, just access to lots of data...

;-)
(sorry that one was too easy)

Show threadggdupontFeb 18, 2025

@fossheim This idea has been in my head all day...

I suggest a new word: "paranousia" which uses similar ancient greek roots with "nous" from "νοῦς" which seems to stand for intelligence/mind.

Another case of #paranousia

Show threadmo7emmedFeb 18, 2025
@fossheim
Science has to be more accessible, expanding beyond the traditional limits of preserving copyrights.
Show threadPalmAndNeedleFeb 18, 2025
@fossheim it is always fascinating to me how stupid smart people can be
Show threadFrank AylwardFeb 18, 2025
@fossheim i noticed the same thing when I asked it questions about my niche field. The answers were worded curiously similar to my own writing style.
Show threadsimon grayFeb 18, 2025
@fossheim That would actually be funny if it wasn’t so depressing
Show threadChantaruFeb 18, 2025
@fossheim it's scary that this person thinks ChatGPT is smart.
Show threadJared White (ResistanceNet ✊)Feb 18, 2025

@fossheim I believe I recall there being a study a while back that showed that the people who are most enthusiastic about this technology are the ones who least understand how it works.

Once you've popped the hood, you find out there is no magic. It's just a zillion Internet-scraped data points in a trenchcoat.

Show threadRobin AdamsFeb 19, 2025

@jaredwhite @fossheim This one:

https://journals.sagepub.com/doi/10.1177/00222429251314491

Show threadJared White (ResistanceNet ✊)Feb 19, 2025
@robinadams @fossheim I appreciate the reference!
Show threadDebbie Goldsmith 🏳️‍⚧️♾️🇺🇦Feb 18, 2025
@fossheim Chatbots are still Eliza; they've just gotten better at fooling people.
Show threadChristine Burns MBE 🏳️‍⚧️📚⧖Feb 18, 2025
@dgoldsmith @fossheim And how does that make you feel?
Show threadLazarou Monkey Terror 🚀💙🌈Feb 18, 2025
@christineburns @dgoldsmith @fossheim lol, I understood that reference!
Show threadDebbie Goldsmith 🏳️‍⚧️♾️🇺🇦Feb 18, 2025
@christineburns @fossheim Tell me more.
Show threadSin VegaFeb 18, 2025
@fossheim I can't believe I put 27 in the bingo machine and it threw out 27 as well! Incredible machine, clearly as intelligent as me
Show threadVirginicusFeb 18, 2025
@fossheim That happened to me, too. The LLM was busted when it repeated an obscure pun that I sneaked into the technical report, just to see if I could get away with it