this post was submitted on 24 Oct 2026
73 points (79.7% liked)
science
29469 readers
462 users here now
A community to post scientific articles, news, and civil discussion.
dart board;; science bs
rule #1: be kind
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Thank you. I appreciate you getting a link to an actual scientific source. There are way too many of these slop sites around these days.
All 7 peer-reviewed PubMed/DOI papers (including Wilder-Smith, Lewis & Pickering 1936, Changizi 2011, and Kareklas 2013) are linked in the References section at the bottom of the article!
Have you tried following any of those so-called "references"? They're all hallucinated.
Eg the first one:
Links to:
https://pubmed.ncbi.nlm.nih.gov/23303785/
If you follow that link you get to:
Polioudakis D, Bhinge AA, Killion PJ, Lee BK, Abell NS, Iyer VR. A Myc-microRNA network promotes exit from quiescence by suppressing the interferon response and cell-cycle arrest genes. Nucleic Acids Res. 2013 Feb 1;41(4):2239-54. doi: 10.1093/nar/gks1452. Epub 2013 Jan 8. PMID: 23303785; PMCID: PMC3575845.
The citation the LLM was failing to hallucinate properly is:
Kareklas K, Nettle D, Smulders TV. Water-induced finger wrinkles improve handling of wet objects. Biol Lett. 2013 Jan 8;9(2):20120999. doi: 10.1098/rsbl.2012.0999. PMID: 23302867; PMCID: PMC3639753.
Absolutely everything a LLM outputs has subtle errors. That's how the technology works, it is essential for it to approximate data so it can merge contexts and output text that "looks about right".
However that's also what makes them unreliable and useless for anything that matters. Verification takes as much or more work than doing the entirety of the work without the LLM.