Another fake disease AI thought was real was "Kyloren syndrome." It showed up in Google's AI Overview.
It was from a hoax paper that @neuroskeptic.bsky.social wrote to expose a predatory journal.
See "The garbage'll do", July 2017, in my hoax anthology, Stinging the Predators:
bit.ly/StingPred
🚨 For nearly two years - and even now, depending on the day and model - AI models could tell you all about bixonimania, a totally made-up disease in a paper funded by “the Professor Sideshow Bob Foundation". My @nature.com story on AI's fallibility in medicine www.nature.com/articles/d41...
Many of the claims made regarding the enhanced efficiency of AI tools are misleading; this one isn't. Large language models (LLMs) are *very* efficient at spreading false information.
#ArtificialIntelligence #LLMs 🧪
www.nature.com/articles/d41...
I am biased but this is the greatest and possibly only Artemis II video you need 🌝 www.youtube.com/shorts/Tc2vN...
My colleague & I teamed up w/ Grounded AI to look for fake/hallucinated references. Our analysis suggests that at least tens of thousands of publications from 2025 likely contain such references. More detz at @nature.com, w/ input from @gcabanac.cpesr.fr, @mhmdhsini.bsky.social & @kowb.bsky.social.
Until you change the incentive structure of academia you will continue to see people fear losing their status and use AI as a way to fast-track productivity bc academia doesn't currently value the meaningful production of knowledge, only the performance of producing it.
This is the inevitable consequence of there being far too many scientists and journals whose only goal is to publish as many papers as possible. Before these papers would have been junk, now they're AI generated junk.
Yeah it’s annoying that such a stupid word has become the term of art in the field.
Tens of thousands of papers published in 2025 contain fake/hallucinated citations, an analysis by Nature and Grounded AI suggests
By @miryamnaddaf.bsky.social and Beth Quill | @nature.com
www.nature.com/articles/d41...
The question about fake references should NOT be “What can be done?”
We KNOW what can be done: human peer reviewers and editors check the damn references individually, by hand.
The question should be, “Why are so many publishers NOT doing those basic checks?”
www.nature.com/articles/d41...
Tens of thousands of papers published in 2025 contain fake/hallucinated citations, an analysis by Nature and Grounded AI suggests
By @miryamnaddaf.bsky.social and Beth Quill | @nature.com
www.nature.com/articles/d41...
I‘d still love a package subscription to the magazines I like rather than subscribing to each individually tho
An unlikely duo — pig semen and eye drops — can halt tumour growth in mice. Could the drops replace needles to treat kids with retinal cancer someday? 👁️🐷
Thanks to Chunxia Zhao at Adelaide University and David Greening @bakerresearchau.bsky.social for chatting with me! www.nature.com/articles/d41...
Interesting counter argument to the ‘break the link’ proposals for electricity and gas prices.
hmm, if the error has a major impact on the paper then why doesn’t it also impact on the work that’s built on the paper?
🧵 New preprint led by @bingbrunton.bsky.social, @elliottabe.bsky.social, @lawrencehu.bsky.social
We gave a worm brain control of a fly body and it walked
What did we learn? Nothing, other than deep reinforcement learning is effective
We call it the digital sphinx
www.biorxiv.org/content/10.6...
For several months, people have been wondering why Reform hasn't declared any crypto donations, despite Nigel Farage saying they'd received some
Now @theobserveruk.bsky.social can reveal why...
This is a granular and informed piece that will teach you about things you don’t know and probably didn’t want to know but which you (we) should know. It also makes some important ancillary points about how homogenized and blindered AI “critique” has become.
Purchasing power in 2026 is $155B for US and $146B for China. Cross over point in 2028 https://www.nature.com/articles/d41586-026-00618-5
The crossover point between US & China investment in R&D may come sooner.
Last year, the White House proposed $5B+ in cuts to NSF, $18B+ in cuts to NIH, and billions more in cuts to NASA, DOE, EPA, NOAA, ...
Hard to stay ahead when support falls behind.
www.nature.com/articles/d41...
NEW: This new report, from BBC, is the best takedown of the insanity of Trump you will see anywhere.
Calm, coherent and repeating only Trump’s own words. Devastating.
Brilliant work once again from Roz Atkins.
(🎥 BBC News)
This is utter garbage. I believe in net zero. We need to get there. But renewables are not competitive. That's why they are subsidised. Fossil fuel markets are highly competitive. It shouldn't be hard to hold these two thoughts at the same time.
You might have formed the impression that Paul Johnson, former director of the Institute for Fiscal Studies, knows what he's talking about. The media treats him like the voice of God. But here he is - and not for the first time - talking out of his hat. Let's break it down in a short thread. 🧵
I'd normally be annoyed at Reuters for doing a fake out where they focus on Rob from Massive Attack, before explaining that Banksy is actually the guy the Daily Mail named years ago, but it's so well written and richly reported I didn't care www.reuters.com/investigates...
This is a pretty interesting unsolved mystery. How did a 2021 social media post linked to a state propaganda campaign know accurate details about SARS-CoV-2 origins that only came to light later? And if the details we know about were right, could the other fairly shocking details also be true?
I think it's just that a perspective based (narrowly?) on jobs created and economic spillovers is going to give a lower return rate than a perspective based on the idea that R&D will in the long term create new technologies driving societal growth. (cf www.newthingsunderthesun.com/pub/ijugr2h6... )
The BEA is a federal agency (part of US Dept of Commerce). It's supposed to be non-partisan objective data on the economy. I'm not aware that they specially updated anything for Trump -- latest multiplier effects in RIMS II Model released April 2025 and this is what UMR are using.
I dunno, but the study says it relies on the Bureau of Economic Analysis' Regional Input-Output Modeling System (RIMS) to spit out the economic effects. The result ($2.57 returned for every $1 spent) is very close to what this group, United for Medical Research, has claimed in previous years.
I'm suing Grammarly over its paid AI feature that presented editing suggestions as if they came from me - and many other writers and journalists - without consent.
State law requires consent before someone's name can be used for commercial purposes.
www.wired.com/story/gramma...
Tbf I wasn’t looking for them.
no sloppelganger for you!
