Every brain tells a story. We need to listen better! #FDG-PET
@davidjonesbrain
Behavioral neurologist and AI director working to end degeneration of mental function and use technology to transform the practice of neurology. Neuro AI Program (NAIP) https://naip.mayo.edu https://www.youtube.com/channel/UCj9q9jnuNZNNZnx48-P1vkg
Every brain tells a story. We need to listen better! #FDG-PET
Cortical oscillations reflect opponent ensemble dynamics through coordinated multifrequency activity https://www.biorxiv.org/content/10.64898/2026.02.20.707132v1
Beyond metaphor: quantitative reconstruction of Waddington's landscape and exploration of cellular behavior academic.oup.com/bib/article/... - really nice review of how big data and new computational methods and reviving an old conceptual framework
Information processing in minds and machines like it as well! Check out our new description of the hippocampus as a Stochastic Latent Oscillatory Diffusion (SLOD) generative engine: bsky.app/profile/davi...
π¨PREPRINT
In our new paper, we link RNNs and neural manifolds by introducing the DDM framework.
We can train networks to embed an arbitary dynamical system in a latent subspace. We illustrate this with simple models of input-driven and autonomous associative memory. Enjoy!
arxiv.org/abs/2602.14885
Our paper is out in @natneuro.nature.com!
www.nature.com/articles/s41...
We develop a geometric theory of how neural populations support generalization across many tasks.
@zuckermanbrain.bsky.social
@flatironinstitute.org
@kempnerinstitute.bsky.social
1/14
Here are individual FDG-PET brain scans from 4 patients with dysexecutive Alzheimerβs disease displayed in StateViewer. Does this look variation along the principal gradient of functional connectivity? Is this biology contributing to your model of Alzheimerβs disease?
Most people can call up pictures in their minds, visualizing the past & summoning images of the future. But for ~4% of us, such mental imagery is weak or absent. New edition of @nature.com has a nice introduction to how research on this phenomenon (aphantasia) opens up novel windows into the brain.π§ͺ
Fascinating neuropsych article from 1919 of a patient with spatial disorientation, with a loss of mental imagery for routes:
watermark02.silverchair.com/archneurpsyc...
We definitely propose neurological mechanisms that fit nicely with GWT! Looking forward to connecting. Here is some related work with me email associated: www.nature.com/articles/s41...
They work for meβ¦ maybe put this in a search engine: βThe Hippocampal Latent Diffusion Engine: A Computational Framework for Memory, Perception, and Cognitive Dysfunctionβ
The brain does that too! doi.org/10.31234/osf...
Bruce Springsteen released a pointedly emotional protest song Wednesday, amid widespread outrage over sweeping immigration raids that have led to violence and fatal shootings at the hands of federal agents.
Block that acetylcholine and just doesn't feel worth it any more...
Cholinergic modulation of dopamine release drives effortful behaviour
www.nature.com/articles/s41...
The hippocampal map has its own attentional control signal!
Our new study reveals that theta #sweeps can be instantly biased towards behaviourally relevant locations. See πΉ in post 4/6 and preprint here π
www.biorxiv.org/content/10.6...
π§΅(1/6)
While AI systems have advanced tremendously, they still lag behind real brains in reliability & efficiency. A new computational unit developed at #FlatironCCN could help close that gap: https://www.simonsfoundation.org/biological-brains-inspire-a-new-building-block-for-artificial-neural-networks
They might be interested in this: bsky.app/profile/drbr...
Thank you! We worked hard on the paper and it ballooned to 11-12k words in the main text before we started cutting it down. My wife is a print and radio journalist and she said βyou have to learn to kill your darlings!β
Thanks. This is helpful. The challenge with this work is that it sits at the intersection of clinical reasoning (neurologic localization), dementia neuroanatomy and comp neuro. There are few audiences that are fluent across all three, which makes a single framing difficult. More attempts to comeβ¦
Seems fitting to use #GenAI to help explain our new model of dementia and the hippocampus as a latent diffusion engine! Here is NotebookLM generated podcast on these topics: open.substack.com/pub/consilie...
I will collect more material on my Substack, but maybe this NotebookLM (AI generated podcast) will give more of an accessible overview of where we are headed: open.substack.com/pub/consilie...
People have been looking at "AI & brains" through the lens of LLMs/transformers. But what latent diffusion? (think text-to-image/video platforms such as Midjourney).
**Class conditioned latent diffusion and semantically cued hippocampus share a remarkably similar computational architecture...**
It bears repeating today: real patriots do not murder people.
#RIPAlexPretti
@simonwheeler.bsky.social here is something new for dementia research, so there is some hope of new pathways coming: bsky.app/profile/davi...
β¦ and the hippocampus? bsky.app/profile/davi...
12/12
This work builds on decades of neuroscience, clinical observation, and modern generative #AI models.
Grateful to collaboratorsβand excited to see where this framework is challenged, refined, or proven wrong.
More discussion to come on my Substack: consilienceofmind.substack.com
11/12 Because SLOD is mechanistic, it generates falsifiable predictions across imaging, electrophysiology, and cognitionβlinking specific dementia phenotypes to failures of compression, timing, priors, and decoding. We highlight some of this in the paper: doi.org/10.31234/osf...
10/12
In SLOD, dementia syndromes arenβt defined by proteinsβbut by where a generative computation breaks.
Tau, Ξ±-syn, TDP-43, and amyloid bias distinct failure modes across compression, timing, priors, decoding, and convergenceβexplaining heterogeneity and mixed pathology.