Gotta love the “or stocks” part. Like Gemini realized its first answer was nuts so it tried to wedge in a better one.
Peter Johnson
@libraryvines.bsky.social
Medical librarian #medlibs also into Tolkien, history, Star Wars, amateur Excel skills, and inclusivity
#medlibs - this was a generated response to a query I have often used as one test to see how the commercially available LLMs are dealing with less straightforward MeSH terms, and the Claude (Sonnet 5) output got it in one. Need to find new test terms 🙂
one way or another the road is going to end on this, and the economic results are likely to be devastating. I don't think AI related technology will go away as a result, but the current landscape isn't going to hold.
Timnit Gebru@timnitgebru.blacksky.app · 7d ago"As Japanese financial newspaper Nikkei Asia found in a recent investigation, just five US tech giants — Alphabet, Microsoft, Amazon, Meta, and Oracle — are hiding an estimated $1.65 trillion in debt that doesn’t appear on balance sheets. futurism.com/artificial-i...
This is just to say I have OCRed the text thet was in tne arch:ve and whicb yov were prabaldy seamhing for "breahtast" Fargme ms fheg mem debo:w5 sb ?mccf mmb 5c oc1b
a permanent challenge with LLMs are their lack of transparency: you can only verify what actions are taken by their outputs. Because of that they are inherently insecure, and should never be used with PII or any item you wouldn't want publicly known. Also applies to workplace provided tools, frankly
TechCrunch@techcrunch.com · last wk.An untold number of Claude chats and Artifacts — the interactive mini apps and documents users can build inside Claude — were found publicly searchable on
Most of the criteria is meant for umbrella reviewers to aid in consistently including higher quality #systematicreviews, but this is also a useful measure for SR planning and evaluation: teams could use this list to identify gaps in their knowledge and planning, and increase the quality of their SRs
Peter Johnson@libraryvines.bsky.social · last wk.#medlibs #evidencesynthesis #systematicreviews www.ovid.com/jnls/jbisrir...
#medlibs #evidencesynthesis #systematicreviews www.ovid.com/jnls/jbisrir...
Consensus operational criteria for including... : JBI Evidence Synthesis
Objective:This international, multidisciplinary study aimed to develop consensus-based operational criteria for umbrella reviewers to apply when...
ovid.com
You guys. You can hold two feelings at once. An adaptation can be a good time and also Not rise to the themes of source material
a core of education is agency, and libraries are at the center of helping our users choose where and how they want to engage or not engage with AI. Especially in lieu of inconsistent governance I’m not surprised that many want to opt out techcrunch.com/2026/07/25/l...
Librarians are hosting viral 'Avoiding AI' workshops for people who are fed up with Big Tech | TechCrunch
At libraries around the country, "Avoiding AI" workshops have elicited unprecedented demand.
techcrunch.com
#medlibs we added a paper by Gagen et al. on development and validation of #searchfilters for #Transgender and #GenderDiverse Populations in #Ovid #MEDLINE and Ovid APA #PsycInfo sites.google.com/a/york.ac.uk...
Fellow professors, how do you have your students address you, and why? Title & last name? title and first name? First name? Something else? #edusky #academicsky #SLISsky
I was today years old when I realized that in addition to showing an unlocked lock, the commonly used #openaccess logo contains both an uppercase "O" and a lowercase "a"
#medlibs - Reflecting on the potential for AI with cataloging and article indexing, I was reminded of my favorite PubMed mis-indexing example.
An interesting consideration of the benefits and limits of agentic AI systems in research. It left me with more questions about effective validation and oversight, but I appreciated the emphasis on the need for shared standards and operating environments media.oregonstate.edu/playlist/ded...
OSU MediaSpace
media.oregonstate.edu
one of the reasons we deaccessioned many of our library's physical titles was due to their age and the potential of containing out-of-date health information and guidance. Scary to think of them as "new" potential sources of AI training data.
Tony Corsentino@notalegalrecord.net · 2w agoAI companies are buying, scanning, and destroying older books to improve their models' training data & avoid problems of "model collapse" (wherein LLMs choke on their own slop). One of their suppliers is Better World Books, which partners with public libraries. www.404media.co/ai-companies...
I was sadly reminded today how glorious Truman State University's logo used to be, referencing both its then new name but also its old name of Northeast Missouri State University by incorporating the weathervane pointing NE on the tower of Kirk Memorial.
#medlibs - I don't agree with all of the characterizations (especially the anthropomorphic comparisons with database searching), but the overview of different categories of prompt design is helpful. doi.org/10.1177/1891...
Sage Journals: Discover world-class research
Subscription and open access journals from Sage, the world's leading independent academic publisher.
doi.org
It’s worth noting that Taco Bell identified the source before the FDA did, and promptly pulled it from their stores. The FDA didn’t even name the source until Taylor Farms outed themselves, and hasn’t issued a recall. Basically, Taco Bell is now a more reliable authority than the FDA.
Forbes@forbes.com · 3w agoTaylor Farms, which supplies major fast food and grocery chains, was identified as the possible source of the cyclosporiasis outbreak Thursday.
Librarians, keep in mind if you're ever using Amazon to reference any book metadata. Perhaps better to rely on OCLC or individual publisher sites.
Six Colors@sixcolors.com · 3w agoUnauthorized AI biographies are all over Amazon (Link)
#medlibs - it's fascinating how far the generative models have come in a brief time. This is not a high quality keyword search, and I would still not trust genAI with controlled vocabulary, but I've seen published reviews with worse searching than this. Made in CoPilot in about 3 minutes.
Automated indexing in MEDLINE (scoping review). Happy to chat about any of the topics. Skip to the good part(s)! #medlibs jmla.mlanet.org/ojs/jmla/art...
"Cartoonification" is now a new favorite word. I think an unresolved issue from the post - the assumption that AI-mediated discovery is only turning up human produced information. Increasingly, it's locating more AI produced content.
Angela Spencer@angspenc.bsky.social · 3w agopart of the problem is seeing human labor as only a cost rather than an investment. www.the-independent.com/tech/ford-ai...
Ford hired AI and sacked humans. It backfired badly
‘We didn’t pay as much attention as we should have to the experience of our most knowledgeable engineers,’ says automaker
the-independent.com
tl;dr within biomedical scholarship, depositing as a preprint tends to be a signifier of better quality than not.
petersuber@petersuber.fediscience.org.ap.brid.gy · 3w agoMajor new study comparing biomedical #preprints with versions later published in peer-reviewed journals: https://www.biorxiv.org/content/10.64898/2026.06.30.735556v1 "The primary claim was unchanged in 39.9% of abstracts, minorly revised in 50.0% [mostly for level of confidence], and […]
I may be the only librarian that some of you know, and that’s delightful but you should also meet more librarians. Most of them are pretty cool.