iiinova — where your digital self keeps evolving

Richard Sever @richardsever.bsky.social · 6h

I prefaced my question by saying it was a single sentence just with lots of semi-colons...

Richard Sever @richardsever.bsky.social · 9h

"Jacobsen’s scenario is a poor map of biological risk...stacking worst-case assumptions about pathogen transmissibility, blurring the distinction between dual-use research and weaponization, and overstating what AI changes" thebulletin.org/2026/10/anni...

thebulletin.org

Annie Jacobsen’s road from Siberia to Armageddon is paved with bad assumptions

Following recent reports of a death at a secretive Siberian disease research lab, author Annie Jacobsen has become a sought-after media interview. Some interviewers are now casting Jacobsen's new best...

0119

Richard Sever @richardsever.bsky.social · 11h

Lanier: "Don't use agents...All AI is is statistical concordance that comes from people. There is no reason to treat it as something with volition. Agent deify AI. Use AI but don't use the agents. That's resistance" #sagebionetworks26

011

Richard Sever @richardsever.bsky.social · 11h

Lanier: "Doctors are now using incredible [AI] tools to help patients...Of course those tools were created by people who want to make us all extinct" #sagebionetworks26

120

Richard Sever @richardsever.bsky.social · 11h

Jaron Lanier: "AI culture features a preponderance of young men...with a greater ease of abstraction than dealing with humans...It is essentially a new religion" #sagebionetworks26

022

Reposted by Richard Sever

Dorothy Bishop @deevybee.bsky.social · 12h

haha. Reminds me of a memorable (apocryphal) meeting of the EPS where the chair responded to a questioner with “if you want to give a talk, you need to get your abstract in on time”

0104

Richard Sever @richardsever.bsky.social · 12h

"A question is a short sentence that ends in a question mark". Conference attendee guidance from @wilbanks.bsky.social at #saegeassembly26 that all should observe

2155

Richard Sever @richardsever.bsky.social · 12h

I'd start by pointing out W&C never claimed to discover that DNA was the genetic material...Their paper was about its structure.

140

Richard Sever @richardsever.bsky.social · 12h

great to be hubristic and wrong!

020

Richard Sever @richardsever.bsky.social · 14h

no - unretracted papers widely believed to be wrong but still cited (mainly by their authors)

010

Richard Sever @richardsever.bsky.social · 14h

the less strict ones exist: Zenodo, Figshare, Vixra. More strict ones would have higher costs and I think then we are really talking about journals...

120

Richard Sever @richardsever.bsky.social · 14h

yeah - publisher-owned preprint servers are a different category. And of course these are primarily intended as previews of content under review they aim to funnel into a journal - as depicted here journals.plos.org/plosbiology/...

journals.plos.org

Biomedical publishing: Past historic, present continuous, future conditional

Academic journals have been publishing the results of biomedical research for more than 350 years. Reviewing their history reveals that the ways in which journals vet submissions have changed over tim...

120

Richard Sever @richardsever.bsky.social · 14h

as happens with journal articles...

100

Richard Sever @richardsever.bsky.social · 14h

I don't think so. I suspect the non-profit servers will converge on similar approaches for human-authored content (where the $ mean they can), but there may be a different universe of unverified/AI content beyond simply Zenodo/Figshare.

100

Richard Sever @richardsever.bsky.social · 15h

Pre-hoc: Volume…and trust. if noise:signal becomes vast the resource uses utility (and costs become unsustainable). Whatever one thinks of the new arXiv policy, that’s basically the reason. Post-hoc: trickier but we want parent institutions to take responsibility for outright fraud

141

Richard Sever @richardsever.bsky.social · 15h

Arsenic life is a good example. No one claims bad intent; they just think it’s wrong. Most people nevertheless thought it should be retracted. Science did retract it but notably one ex-Science editor argued they should not have.

020

Richard Sever @richardsever.bsky.social · 15h

Yes - otoh we do retract honestly created papers where deep flaws emerge afterwards. Sometimes the authors concede those; sometimes they don’t.

120

Richard Sever @richardsever.bsky.social · 15h

Yeah - I also think it doesn’t work when it comes to things like ethical issues. These can be a nightmare, take vast amounts of staff time and need a designated third-party arbiter. Very underappreciated (by academics) part of the process.

110

Richard Sever @richardsever.bsky.social · 08/10/2026

But it’s a really important thing to think about in the context of proposals for publishing reform. Publish-Review-Curate and ‘preprint-only’ policies are currently pushed by various funders, but what do retractions look like in these workflows and who arbitrates them?

2103

Richard Sever @richardsever.bsky.social · 08/10/2026

2) Does intent make a difference? 3) is it different if the venue makes some claim to have verified the content (biorxiv is not a journal, would someone asked the same question of Zenodo?, etc. 4) Does the scale of the problem change things?

170

Richard Sever @richardsever.bsky.social · 08/10/2026

There are multiple strands here. 1) the general question of how we as a community handle work judged wrong (do we retract Pauling’s paper on DNA structure - is there a statute of limitations? How is the water memory paper still out there? Was it right to retract As life?)

171

Richard Sever @richardsever.bsky.social · 08/10/2026

The question is whether this should be any different from preprints fabricated without AI…

1164

Richard Sever @richardsever.bsky.social · 08/10/2026

And finally any version of “Do you know who I am?”. Really no answer to that benefits you in the face of a rejection, and worst-case scenario the retort will be “yeah and you’re lucky we didn’t factor that in to the decision”. 🤣

2162

Richard Sever @richardsever.bsky.social · 08/10/2026

Other strategies not recommended: citing work by Galileo, Prusiner or Watson & Crick as comparable scenarios.

371

Richard Sever @richardsever.bsky.social · 08/10/2026

“This guy’s cursing me up. I better reconsider my decision”, said no editor ever…

2226

Richard Sever @richardsever.bsky.social · 07/10/2026

"OA the warm-up act for AI...Most authors don't read the small print...don't realize what they're giving away". Science & Nature editors on the drawbacks of CC BY licenses. I don't agree with all they say but have always said we need more debate and less dogmatism on this youtu.be/Q04PAohvSYU?...

youtu.be

Who Did the Analysis? Editors on Disclosure and Peer Review in the Age of AI Agents

YouTube video by Broad Institute

032

Richard Sever @richardsever.bsky.social · 07/10/2026

Nature E-i-C @magdalenaskipper.bsky.social: "peer reviewers are much harsher critics...Editors are more likely to say let's publish it [as is]" Important in the context of accusations it's "journals" demanding more experiments of authors. youtu.be/Q04PAohvSYU?...

youtu.be

Who Did the Analysis? Editors on Disclosure and Peer Review in the Age of AI Agents

YouTube video by Broad Institute

186

Reposted by Richard Sever

Victoria Strauss @victoriastrauss.com · 07/10/2026

Canada is refusing to issue ISBNs to books primarily generated with AI. This illustrates both the urgency of the problem of the growing flood of slop books, and also the toothlessness of addressing it on the honor system without there also being consequences for lying

Use of artificial intelligence
By registering with ISBN Canada, you declare that your book was created by a human author. You may use AI tools only for limited support tasks, such as brainstorming ideas, checking grammar, or formatting your book. You must not use AI to generate the text, create the artwork, or translate the book. ISBN Canada does not issue ISBNs for works whose content is primarily generated by AI. LAC is currently reviewing its policy regarding the use of AI content creation and our guidance is subject to change.

252610683

Reposted by Richard Sever

John Ross @johnrossmd.bsky.social · 07/10/2026

The annual curve of flu cases is starting to rise sharply in the US. This would be a great week to get vaccinated, if you haven't already

Influenza Positive Tests Reported to CDC by Clinical Laboratories,
National Summary, 2025-26 Season, week ending Sep 26, 2026. From: https://www.cdc.gov/fluview/surveillance/2026-week-38.html

14120

Reposted by Richard Sever

Nathaniel Comfort @nccomfort.bsky.social · 06/10/2026

Reposting for the morning crowd my mini-thread from yesterday on the new Sponsel & Sutton paper on Rosalind Franklin's work. tl;dr: Why didn't Franklin find the double helix? Because she wasn't looking for it! She had her own DNA project that had nothing to do w/ the gene.

1108

Richard Sever @richardsever.bsky.social · 06/10/2026

And TMV. Is it safe to say she might have got a Nobel for that had she not died before it was awarded?

010

Richard Sever @richardsever.bsky.social · 05/10/2026

this should be fixed now - let us know if you are still having problems.

110

Richard Sever @richardsever.bsky.social · 05/10/2026

It’s not even a mean as the numerator and denominator are from different data sets 🤪

010

Reposted by Richard Sever

Juan Nunez-Iglesias @jni.codes · 05/10/2026

You may be horrified to learn that at Monash until recently (not sure if still true or not) one of the measures was, I shit you not, “Cumulative Impact Factor”, which is the *sum* of the IFs of the journal of each of your papers.

7264

Reposted by Richard Sever

Benoit Bruneau @benoitbruneau.bsky.social · 04/10/2026

Happy Pluripotency Day! #Oct4

14916

Richard Sever @richardsever.bsky.social · 04/10/2026

Interestingly Claude could only identify one and it didn’t have many reviews…

000

Richard Sever @richardsever.bsky.social · 04/10/2026

Wanted: email client plug-in that filters AI-generated emails to a separate inbox I pay almost as little attention to as the spam folder...

2200

Richard Sever @richardsever.bsky.social · 04/10/2026

I use both. Pangram tends to have more false negatives and GPTzero more false positives IME but not always. The browser plug is handy though

010

Richard Sever @richardsever.bsky.social · 04/10/2026

Yeah but I feel the impact on math is more radical. Fundamental change in a fundamental type of knowledge generation.

110

Richard Sever @richardsever.bsky.social · 03/10/2026

And your regular reminder of the innumeracy of this metric (a pseudo mean calculated from two different datasets and applied to a non-normal distribution) www.biorxiv.org/content/10.1...

34311

Richard Sever @richardsever.bsky.social · 03/10/2026

Any sane academic evaluation policy would forbid mention of Impact Factor

67821

Richard Sever @richardsever.bsky.social · 03/10/2026

yes - this to me is a massive question. no user is really faced with the true cost. I've seen estimates for how much a chatbot query actually costs and they are shocking. The response is usually "tech costs will come down", but this is energy costs and if we could have solved them....

1101

Richard Sever @richardsever.bsky.social · 03/10/2026

you never know. I was very impressed with Elham Azizi youtu.be/PzsUUbiliH0?...

youtu.be

Interview with: Elham Azizi

YouTube video by CSHL Leading Strand

100

Reposted by Richard Sever

Richard Sever @richardsever.bsky.social · 03/10/2026

GPTzero has a good chrome plug-in. I am relieved to see it identify all my Gdocs as 100% human

231