Lars Lau Raket

My paper on statistical disease progression modeling in Alzheimer's disease has just been published in @FrontAIBigData. You can find it here doi.org/10.3389/fdata.… Want to try it out on your own data? Code can be found here github.com/larslau/progmod

Why does this matter? Knowing when biomarkers change helps us stage disease and optimize timing for treatments—crucial for trials & patients. (6/6) 🙏 Thanks to all collaborators! @NicholasAshton @SebastianPalmqv @OskarHansson9 @biofinder_study

Now, BioFINDER. The patterns are strikingly consistent—timing of biomarker changes aligns closely with ADNI. (5/6)

🧠🧪✨Can we map the full timeline of biomarker changes in Alzheimer’s disease—from the earliest shifts (a decade before symptoms) to advanced dementia? Our new Brain paper says: Yes! Two large cohorts, one consistent story. (1/6) 🔗doi.org/10.1093/brain/…

@Plinz Horseshoe crabs and scorpions have eyes around their bodies, but do they feel they are looking out of them? Maybe they feel they’re looking out of their front eyes with excellent peripheral vision? Or maybe they have a single omnidirectional perspective anchored in their body?

@Plinz The perspective from my two front-mounted photosensors minimizes prediction error. But there must also be a computational constraints on how perspective is placed?

It's tempting to think that the subjective visual (ego centric) perspective has to be anchored at your retinas, and you are basically always seeing what our eyes are looking at, but out of body experiences demonstrate otherwise.

“The arc of history is long, and it bends toward” *flips through history book* Oh. Oh no. Oh no no no

AI psychosis psychosis is rampant now

New paper & surprising result. LLMs transmit traits to other models via hidden signals in data. Datasets consisting only of 3-digit numbers can transmit a love for owls, or evil tendencies. 🧵

Our most important paper, out now! A massive international collaboration, led by @AlexisMoscoso9, featuring over 6500 visually assessed tau PET scans covering the #Alzheimers spectrum! Thank you very much to the numerous contributors and collaborators who made this gem possible.

Strong title:

There is now something new under the sun: you can literally experience a completely new perceptual color that no one has ever seen before

biology is nothing but trouble

Always my favourite post at the beginning of the year: Registration for our course “Biomarkers for neurodegenerative diseases” is now open! This year’s edition takes place at @UCLIoN in central London on May 12-16. biomarker-course.org

Excited to share the REAL AD design paper, now out in @alzdemjournals The study has been live for 5 months, and we are elated and grateful that 5000 participants have already enrolled! Stay turned for first results! alz-journals.onlinelibrary.wiley.com/doi/10.1002/al…

A friendly reminder that our @ISTAART @DesignDataPIA All Member Meeting is happening tomorrow (Tuesday 09/10) at 2 p.m. CT! Join us if you can 😎 @LarsRaket @DrAnaCapuano @mikemalekahmadi @natggoncalves @JetMJVonk @ni_corv

@Plinz Agree. The point I tried to get across was that while there can exist things that are not learnable, I highly doubt that things can feel non-learnable (= impossible to model) since the awareness of the thing means that one has already built a model that identifies the thing.

@LarsRaket Most things are not learnable, because they cannot be predicted from the available observables, stored data and discoverable models

Why is the world so learnable? I think that’s because all complex structure is the result of natural control systems. Since control depends on implementing a model of what’s being controlled (Good Regulator theorem), models behind natural structure are efficiently discoverable.

@Plinz It is likely that things that cannot be modeled will not be observable to the agents at all. I cannot imagine an observable phenomenon that does not appear learnable. Perhaps you have a good example in mind, @Plinz? 4/4

@Plinz The reason is that successful agents recognize that their models are predictive, so they will focus on the things in their world that make sense. Things that cannot be modeled will be deemed uninteresting chaos/noise (that can be described as such). 3/4