Andreas Longva
@andlon
Previously: Researcher and PhD candidate @ RWTH Aachen University, now probably disabled by Long COVID. Physics simulation, mathematics, software and science andreaslongva.com
Google AI still doing about 50/50 on basic facts in my experience. Kinda crazy that they just don't care. I see way too many people using Google AI overviews for fact checking. Please don't! Go to trusted sources instead. The skills to verify information have never been more important.
This is the weirdest timeline. The Jacobian Conjecture was possibly just proven to be false by a guy prompting an LLM and publishing a valid counterexample in a half-assed X post. Open problem since 1884/1939, one of the most known open problems in math. 🤯
I think they're just paraphrasing the article (see snippet). Anxiety is particularly relevant because post-viral panic/anxiety disorder is a frequent presentation of PAIS, so it could be that the observed differences are explained partially by a subset of patients with anxiety.
Irriterer meg over faktafeil i Live Landmarks nyeste debattinnlegg på forskersonen.no. Hun påstår at bare 17 av hundrevis av personer med ME-diagnose endte opp med en reell ME-diagnose etter medisinsk undersøkelse. Dette er helt feil (se bilde). Det var bare 4 stk som endte opp med annen diagnose,
Came across two really great talks on the Finite Element Method (FEM) and Isogeometric Analysis (IGA) by Thomas J. R. Hughes (father of IGA). Although very technical talks, I'm sure many might enjoy this particular slide at the end of one of the talks! (Links below)
Thanks! From a quick skim, they unfortunately seem to conflate PEM and exercise intolerance, see e.g. attached screenshot. Unfortunate that researchers keep using these terms wrong, it really leads to a lot of confusion...
Hmm, that's not how I interpret their text, but you may be right. No test is perfect, that's definitely true! But in the re-analysis of the data from the above article, they discuss sensitivity and specificity. I've included one of their plots of the relative change for every patient and control
Thank you for the response! I may be mistaken, but I believe those statements must be understood at a group level, e.g. the patients as an aggregate tend to exhibit more decline on day 2 than controls. Note that the authors themselves don't seem to consider 2-day CPET as a biomarker (see image)
My concern with the hand strength measurement is that it might not be particularly specific, and could potentially be associated with fatigue in general rather than ME/CFS specifically. See for example this study, where cancer-related fatigue patients exhibit similar results as ME patients
Btw, ChatGPT 5 Thinking could solve these problems on release day. I think the person there must have used the Auto or Fast versions. I could never reproduce any of these trip-ups, as for me it always solved them correctly. See also attached pictures for what is presumably a novel problem ...
TIL if you google "comic sans" it'll display the results in comic sans 😅
@aftenposten.no: hvorfor fortsetter dere å publisere slike artikler uten i det hele tatt å nevne at Russland som regel alltid lyver? De påstår gang på gang at de skyter ned droner som åpenbart treffer målene sine. De har jo *null* kredibilitet, det burde dere opplyse om!
I wanted to try to have ChatGPT help me in dictating a blog post about computer graphics research using voice mode. I had an intro and we got started on the next paragraphs, until... ChatGPT refused to go further, saying that its guidelines prevent it from talking about it! What the hell?
Made the mistake of opening up X in my browser today. 1 Notification: A post by Elon Musk, who I *don't* follow. I'm glad most of the people I engaged with moved to Bluesky.
Ikke hos meg, dessverre. Kanskje det tar litt tid før det oppdateres hos alle?
Forøvrig får jeg Invalid Handle for @Aftenposten.no, jeg vet ikke om dette er et kjent problem også for andre eller om det bare er for meg @aftarkiv.bsky.social
I've been using ChatGPT more for everyday tasks. It's become remarkably good at image recognition tasks, like here where it figures out the manufacturer of a shard of porcelain
I haven't read the whole paper but I'm really confused. This paper seems to pretty much confirm 2-day CPET as a PEM marker, doesn't it? I haven't yet found the stats you mentioned... Would have to read more thoroughly tomorrow
Thought I'd share an example for my paradis #rustlang crate with the Bluesky crowd. It lets you (often safely) express parallel patterns that are really hard to do in Rust without the abstractions provided by paradis. Check out my blog post for more! andreaslongva.com/blog/enter-p...