this post was submitted on 26 Oct 2024
1149 points (97.3% liked)

Science Memes

10897 readers
1329 users here now

Welcome to c/science_memes @ Mander.xyz!

A place for majestic STEMLORD peacocking, as well as memes about the realities of working in a lab.



Rules

  1. Don't throw mud. Behave like an intellectual and remember the human.
  2. Keep it rooted (on topic).
  3. No spam.
  4. Infographics welcome, get schooled.

This is a science community. We use the Dawkins definition of meme.



Research Committee

Other Mander Communities

Science and Research

Biology and Life Sciences

Physical Sciences

Humanities and Social Sciences

Practical and Applied Sciences

Memes

Miscellaneous

founded 2 years ago
MODERATORS
 
you are viewing a single comment's thread
view the rest of the comments
[–] Sabre363@sh.itjust.works 77 points 1 week ago (5 children)

Easily by thwarted by simply proofreading your shit before you submit it

[–] yamanii@lemmy.world 77 points 1 week ago (1 children)

There are professional cheaters and there are lazy ones, this is gonna get the lazy ones.

[–] MalditoBarbudo@programming.dev 27 points 1 week ago

I wouldn't call "professional cheaters" to the students that carefully proofread the output. People using chatgpt and proofreading content and bibliography later are using it as a tool, like any other (Wikipedia, related papers...), so they are not cheating. This hack is intended for the real cheaters, the ones that feed chatgpt with the assignment and return whatever hallucination it gives to you without checking anything else.

[–] xantoxis@lemmy.world 72 points 1 week ago* (last edited 1 week ago) (1 children)

Is it? If ChatGPT wrote your paper, why would citations of the work of Frankie Hawkes raise any red flags unless you happened to see this specific tweet? You'd just see ChatGPT filled in some research by someone you hadn't heard of. Whatever, turn it in. Proofreading anything you turn in is obviously a good idea, but it's not going to reveal that you fell into a trap here.

If you went so far as to learn who Frankie Hawkes is supposed to be, you'd probably find out he's irrelevant to this course of study and doesn't have any citeable works on the subject. But then, if you were doing that work, you aren't using ChatGPT in the first place. And that goes well beyond "proofreading".

[–] And009@reddthat.com 16 points 1 week ago

This should be okay to do. Understanding and being able to process information is foundational

[–] abbadon420@lemm.ee 25 points 1 week ago

But that's fine than. That shows that you at least know enough about the topic to realise that those topics should not belong there. Otherwise you could proofread and see nothing wrong with the references

[–] Sylvartas@lemmy.world 9 points 1 week ago (1 children)

Bold of you to assume students proofread what chatGPT spits out

[–] Sabre363@sh.itjust.works 5 points 1 week ago

I've worked as tutor, I know those little idiots ain't proofing a got-damn thing

[–] psud@aussie.zone 7 points 1 week ago* (last edited 1 week ago) (4 children)

LLMs can't cite. They don't know what a citation is other than a collection of text of a specific style

You'd be lucky if the number of references equalled the number of referenced items even if you were lucky enough to get real sources out of an LLM

If the student is clever enough to remove the trap reference, the fact that the other references won't be in the University library should be enough to sink the paper

[–] auzy@lemmy.world 17 points 1 week ago* (last edited 1 week ago)

They can. There was that court case where the cases cited were made up by chatgpt. Upon investigation it was discovered it was all hallucinated by chatgpt and the lawyer got into deep crap

[–] uis@lemm.ee 9 points 1 week ago* (last edited 1 week ago) (1 children)

LLMs can't cite. They don't know what a citation is other than a collection of text of a specific style

LLMs can cite. It's called Retrival-Augmented Generation. Basically LLM that can do Information Retrival, which is just academic term for search engines.

You'd be lucky if the number of references equalled the number of referenced items even if you were lucky enough to get real sources out of an LLM

You can just print retrival logs into references. Well, kinda stretching definition of "just".

[–] notthebees@reddthat.com 1 points 1 week ago (2 children)

My question is that the thing they are citing actually exists and if it does exist, contains the information it claims.

[–] uis@lemm.ee 1 points 1 week ago* (last edited 1 week ago)

the thing they are citing actually exists

In case of RAGs it exists in searched dataset.

and if it does exist, contains the information it claims.

Not guaranteed.

[–] FutileRecipe@lemmy.world 1 points 1 week ago

Depends. In my experience, it usually does exist. Now there are hallucinations where GPT makes up stuff or just misinterprets what it read. But it's super easy to read the GPT output, look at the cited work, skim works for relevance, then tweak the wording and citing to match.

If you just copy/paste and take GPT's word for it without the minimal amount of checking, you're digging your own grave.

[–] interdimensionalmeme@lemmy.ml 3 points 1 week ago

Let's give that a try, this is classic model, no retrieval

Chatgpt how deadly is covid19, two paragraph answer, include citations


COVID-19 varies in severity, with mortality rates heavily influenced by factors such as age, underlying health conditions, and access to healthcare. The World Health Organization (WHO) estimates a global case fatality rate (CFR) of around 1-3%, though this can be much higher among older adults and immunocompromised individuals (WHO, 2023). In comparison to seasonal influenza, COVID-19 has a higher mortality rate, particularly during the initial outbreaks before widespread vaccination. Long-term complications, often referred to as "long COVID," further increase the disease's burden beyond acute infection (NIH, 2022).

Vaccination efforts, antiviral treatments, and improved healthcare protocols have reduced mortality rates over time, especially in developed countries (CDC, 2023). However, variants of concern, such as Delta and Omicron, have shown differing degrees of transmissibility and immune escape, complicating pandemic management. Even with lowered death rates in certain populations, COVID-19 remains a significant public health threat, particularly in areas with low vaccination coverage or overstretched healthcare systems (Johns Hopkins University, 2023).

References:

World Health Organization (WHO). (2023). COVID-19 Mortality Update.

Centers for Disease Control and Prevention (CDC). (2023). Impact of Vaccines on COVID-19 Severity.

National Institutes of Health (NIH). (2022). Long COVID and Post-Acute Sequelae of SARS-CoV-2 Infection.

Johns Hopkins University. (2023). COVID-19 Global Trends and Variants.
[–] TachyonTele@lemm.ee 0 points 1 week ago* (last edited 1 week ago) (1 children)

Duckduckgos llm cites it's sources. I've seen screenshots of googles llm doing the same.

You're basing your entire comment on made up nonsense, that hilariously might as well be a llm result.

[–] Swedneck@discuss.tchncs.de 2 points 1 week ago

i've tried to get their ai thing to cite and it does the standard AI thing of generating a sensible-looking output that isn't actually valid, it'll link to pages that don't exist but that do lead to valid websites.