Dan Levenstein
@dlevenstein.bsky.social
Neuroscientist, in theory.
Studying sleep and navigation in 🧠s and 💻s.
Wu Tsai Investigator, Assistant Professor of Neuroscience at Yale.
An emergent property of a few billion neurons, their interactions with each other and the world over ~1 century.
Studying sleep and navigation in 🧠s and 💻s.
Wu Tsai Investigator, Assistant Professor of Neuroscience at Yale.
An emergent property of a few billion neurons, their interactions with each other and the world over ~1 century.
Pinned
Dan Levenstein
@dlevenstein.bsky.social
· Jun 23
Thrilled to announce I'll be starting my own neuro-theory lab, as an Assistant Professor at @yaleneuro.bsky.social @wutsaiyale.bsky.social this Fall!
My group will study offline learning in the sleeping brain: how neural activity self-organizes during sleep and the computations it performs. 🧵
My group will study offline learning in the sleeping brain: how neural activity self-organizes during sleep and the computations it performs. 🧵
Reposted by Dan Levenstein
I am really proud that eLife have published this paper. It is a very nice paper, but you need to also read the reviews to understand why! 1/n
"The inevitability and superfluousness of cell types in spatial cognition". Intuitive cell types are found in random artificial networks using the same selection criteria neuroscientists use with actual data. elifesciences.org/reviewed-pre... 1/2
elifesciences.org
November 25, 2025 at 8:34 PM
I am really proud that eLife have published this paper. It is a very nice paper, but you need to also read the reviews to understand why! 1/n
Reposted by Dan Levenstein
Y’all are reading this paper in the wrong way.
We love to trash dominant hypothesis, but we need to look for evidence against the manifold hypothesis elsewhere:
This elegant work doesn't show neural dynamics are high D, nor that we should stop using PCA
It’s quite the opposite!
(thread)
We love to trash dominant hypothesis, but we need to look for evidence against the manifold hypothesis elsewhere:
This elegant work doesn't show neural dynamics are high D, nor that we should stop using PCA
It’s quite the opposite!
(thread)
“Our findings challenge the conventional focus on low-dimensional coding subspaces as a sufficient framework for understanding neural computations, demonstrating that dimensions previously considered task-irrelevant and accounting for little variance can have a critical role in driving behavior.”
Neural dynamics outside task-coding dimensions drive decision trajectories through transient amplification
Most behaviors involve neural dynamics in high-dimensional activity spaces. A common approach is to extract dimensions that capture task-related variability, such as those separating stimuli or choice...
www.biorxiv.org
November 25, 2025 at 4:16 PM
Y’all are reading this paper in the wrong way.
We love to trash dominant hypothesis, but we need to look for evidence against the manifold hypothesis elsewhere:
This elegant work doesn't show neural dynamics are high D, nor that we should stop using PCA
It’s quite the opposite!
(thread)
We love to trash dominant hypothesis, but we need to look for evidence against the manifold hypothesis elsewhere:
This elegant work doesn't show neural dynamics are high D, nor that we should stop using PCA
It’s quite the opposite!
(thread)
There are a small number of papers that I still think about regularly 10 years after reading them and this is one of them.
Izbikevich also had an extremely cool, wild, and thought-provoking model of how this might work that I feel never really got fully fleshed out.
www.izhikevich.org/publications...
www.izhikevich.org/publications...
www.izhikevich.org
November 24, 2025 at 1:54 PM
There are a small number of papers that I still think about regularly 10 years after reading them and this is one of them.
1) This is a great idea.
2) I would be happy to mentor a project like this. If you want to do a project eg at the intersection of Philosophy of Science and NeuroAI, please feel free to reach out to discuss, or just put my name as a suggested mentor!
2) I would be happy to mentor a project like this. If you want to do a project eg at the intersection of Philosophy of Science and NeuroAI, please feel free to reach out to discuss, or just put my name as a suggested mentor!
🗣️ With the support of the @danafoundation.bsky.social, we are very excited to announce the Philosophy & Neuroscience Collaborative Mentorship Program! 🧠
For more details & submission requirements, visit: philandneuro.com/mentorship
(This is 1/2 announcements we will make over the next 1-2 weeks.)
For more details & submission requirements, visit: philandneuro.com/mentorship
(This is 1/2 announcements we will make over the next 1-2 weeks.)
November 23, 2025 at 7:12 PM
1) This is a great idea.
2) I would be happy to mentor a project like this. If you want to do a project eg at the intersection of Philosophy of Science and NeuroAI, please feel free to reach out to discuss, or just put my name as a suggested mentor!
2) I would be happy to mentor a project like this. If you want to do a project eg at the intersection of Philosophy of Science and NeuroAI, please feel free to reach out to discuss, or just put my name as a suggested mentor!
“Our findings challenge the conventional focus on low-dimensional coding subspaces as a sufficient framework for understanding neural computations, demonstrating that dimensions previously considered task-irrelevant and accounting for little variance can have a critical role in driving behavior.”
Neural dynamics outside task-coding dimensions drive decision trajectories through transient amplification
Most behaviors involve neural dynamics in high-dimensional activity spaces. A common approach is to extract dimensions that capture task-related variability, such as those separating stimuli or choice...
www.biorxiv.org
November 23, 2025 at 1:38 PM
“Our findings challenge the conventional focus on low-dimensional coding subspaces as a sufficient framework for understanding neural computations, demonstrating that dimensions previously considered task-irrelevant and accounting for little variance can have a critical role in driving behavior.”
Reposted by Dan Levenstein
Most neuroscientists would agree that the brain is far more sophisticated than ANNs - so why would we force tools and interpretations that are too trivial to explain ANNs today, much less the brain? We wouldn't.
November 20, 2025 at 2:42 PM
Most neuroscientists would agree that the brain is far more sophisticated than ANNs - so why would we force tools and interpretations that are too trivial to explain ANNs today, much less the brain? We wouldn't.
Reposted by Dan Levenstein
We're almost at the end of the year, and that means an end-of-year review! Send me your favorite NeuroAI papers of the year (preprints or published, late last year is fine too).
November 19, 2025 at 4:14 PM
We're almost at the end of the year, and that means an end-of-year review! Send me your favorite NeuroAI papers of the year (preprints or published, late last year is fine too).
Reposted by Dan Levenstein
What are the top 3 dream experiments you'd like neurophysiologists to do, that would help your modeling the most?
November 17, 2025 at 2:34 AM
What are the top 3 dream experiments you'd like neurophysiologists to do, that would help your modeling the most?
Since so many people are here for #sfn25 this week, maybe a good time to mention the algorithmic “For You” feed:
bsky.app/profile/did:...
which shows you posts that people who liked the same things as you liked 🫧📌
bsky.app/profile/did:...
which shows you posts that people who liked the same things as you liked 🫧📌
Welcome to the ✨For You✨ feed!
It finds people who liked the same posts as you, and shows you what else they've liked recently.
📌 Pin to add it to your top bar
❤️ Like the feed and repost to spread the goodness
It finds people who liked the same posts as you, and shows you what else they've liked recently.
📌 Pin to add it to your top bar
❤️ Like the feed and repost to spread the goodness
November 18, 2025 at 2:43 PM
Since so many people are here for #sfn25 this week, maybe a good time to mention the algorithmic “For You” feed:
bsky.app/profile/did:...
which shows you posts that people who liked the same things as you liked 🫧📌
bsky.app/profile/did:...
which shows you posts that people who liked the same things as you liked 🫧📌
Will be presenting a poster on @apeyrache.bsky.social row, Tues AM, with an update our sequential predictive learning preprint:
-CA3-like (sparse lognormal) connectivity gives a hippocampus-like (orthogonalized) map
-non-spatial representation! splitters, lap/time cells, action plan…we got it all🤑
-CA3-like (sparse lognormal) connectivity gives a hippocampus-like (orthogonalized) map
-non-spatial representation! splitters, lap/time cells, action plan…we got it all🤑
TFW you land in San Diego and the bsky is hopping #sfn25
November 15, 2025 at 5:02 AM
Will be presenting a poster on @apeyrache.bsky.social row, Tues AM, with an update our sequential predictive learning preprint:
-CA3-like (sparse lognormal) connectivity gives a hippocampus-like (orthogonalized) map
-non-spatial representation! splitters, lap/time cells, action plan…we got it all🤑
-CA3-like (sparse lognormal) connectivity gives a hippocampus-like (orthogonalized) map
-non-spatial representation! splitters, lap/time cells, action plan…we got it all🤑
“and an otherwise normal life to lead”
November 15, 2025 at 4:53 AM
“and an otherwise normal life to lead”
Somehow clicked the "please send me piles of email spam" checkbox when registering for SfN this year and boy do I have regrets.
November 11, 2025 at 7:08 PM
Somehow clicked the "please send me piles of email spam" checkbox when registering for SfN this year and boy do I have regrets.
Reposted by Dan Levenstein
It’s possible to see social media, and now AI, as the new radio — the new information technology which will ruin democracies around the globe if we don’t find a way to prise off billionaires’ control of those new info channels.
November 9, 2025 at 2:56 PM
It’s possible to see social media, and now AI, as the new radio — the new information technology which will ruin democracies around the globe if we don’t find a way to prise off billionaires’ control of those new info channels.
A loss of trust in information is just as bad as a loss of trustworthy information.
I spotted this on Mastodon and I find it horrible, not least for the speed with which this has happened.
November 9, 2025 at 2:40 PM
A loss of trust in information is just as bad as a loss of trustworthy information.
Starting to wonder if any of us will be at sfn or if we’re all going to be hanging out in airports for the week…
Really disappointing that our #NIH colleagues will NOT be at @sfn.org this year, will NOT be discussing science, will NOT be advising us on grants, will NOT be sharing results or advancing research. 🤐
Sigh. NIH normally sends several hundred scientists to the SFN annual meeting to learn, exchange info, come up with new ideas, and advance science. (The exchange of ideas is the very core of the scientific enterprise.)
This year, no one from NIH will attend due to the gov't implosion.
This year, no one from NIH will attend due to the gov't implosion.
November 8, 2025 at 12:10 AM
Starting to wonder if any of us will be at sfn or if we’re all going to be hanging out in airports for the week…
Reposted by Dan Levenstein
I'm not saying we don't have "systems that rival human intelligence in key tasks" (though "key" is doing some heavy lifting). I'm saying that if you're going to make this your definition of AGI, you've been taking the piss all along.
November 6, 2025 at 9:16 PM
I'm not saying we don't have "systems that rival human intelligence in key tasks" (though "key" is doing some heavy lifting). I'm saying that if you're going to make this your definition of AGI, you've been taking the piss all along.
Not necessarily neuroscience, but I once heard introducing+defining the MDP formalism is the “RL handshake”
What other seemingly obligatory phrases do you notice in neuro papers?
November 4, 2025 at 9:17 PM
Not necessarily neuroscience, but I once heard introducing+defining the MDP formalism is the “RL handshake”
My favorite part of pragmatism is when it’s like “maybe instead of worrying about shit that doesn’t matter we should worry about shit that does.”
“Oh and btw we’ll learn a lot more about the shit that doesn’t in the process anyway.” 🫣
“Oh and btw we’ll learn a lot more about the shit that doesn’t in the process anyway.” 🫣
[2/9] We argue that instead of getting stuck on metaphysical debates (is AI conscious?), we should treat personhood as a flexible bundle of obligations (rights & responsibilities) that societies confer.
November 3, 2025 at 12:54 PM
My favorite part of pragmatism is when it’s like “maybe instead of worrying about shit that doesn’t matter we should worry about shit that does.”
“Oh and btw we’ll learn a lot more about the shit that doesn’t in the process anyway.” 🫣
“Oh and btw we’ll learn a lot more about the shit that doesn’t in the process anyway.” 🫣
Lyrics are just a vehicle for syllables.
November 1, 2025 at 9:06 PM
Lyrics are just a vehicle for syllables.
Reposted by Dan Levenstein
[2/9] We argue that instead of getting stuck on metaphysical debates (is AI conscious?), we should treat personhood as a flexible bundle of obligations (rights & responsibilities) that societies confer.
October 31, 2025 at 12:33 PM
[2/9] We argue that instead of getting stuck on metaphysical debates (is AI conscious?), we should treat personhood as a flexible bundle of obligations (rights & responsibilities) that societies confer.
Reposted by Dan Levenstein
New on the Archive:
Stump, David J. (2025) Lessons from Pragmatism for Philosophers of Science: Nine Teachings and a Cautionary Tale. [Preprint]
https://philsci-archive.pitt.edu/27072/
Stump, David J. (2025) Lessons from Pragmatism for Philosophers of Science: Nine Teachings and a Cautionary Tale. [Preprint]
https://philsci-archive.pitt.edu/27072/
November 1, 2025 at 2:11 PM
New on the Archive:
Stump, David J. (2025) Lessons from Pragmatism for Philosophers of Science: Nine Teachings and a Cautionary Tale. [Preprint]
https://philsci-archive.pitt.edu/27072/
Stump, David J. (2025) Lessons from Pragmatism for Philosophers of Science: Nine Teachings and a Cautionary Tale. [Preprint]
https://philsci-archive.pitt.edu/27072/
Really hoping bifurcations are the new manifolds. What a time to be alive 🥲
My co-authors have yet to move to Bluesky, so I'm pleased to announce our latest work has just been published in @nature.com Neuroscience. Amazing work led by Junheng Li, revealing that falling asleep follows a predictable bifurcation pattern #neuroskyence #sleep
www.nature.com/articles/s41...
www.nature.com/articles/s41...
Falling asleep follows a predictable bifurcation dynamic - Nature Neuroscience
Li et al. propose a conceptual framework to study the phenomenon of falling asleep based on electroencephalogram data. They show that a tipping point marks the brain’s nonlinear wake-to-sleep transiti...
www.nature.com
October 29, 2025 at 1:11 AM
Really hoping bifurcations are the new manifolds. What a time to be alive 🥲
Reposted by Dan Levenstein
"Because science rejects claims to truth based on authority and depends on the criticism of established ideas, it is the enemy of autocracy. Because scientific knowledge is tentative and provisional, it is the enemy of dogma. "
October 25, 2025 at 9:41 PM
"Because science rejects claims to truth based on authority and depends on the criticism of established ideas, it is the enemy of autocracy. Because scientific knowledge is tentative and provisional, it is the enemy of dogma. "
Reposted by Dan Levenstein
George Box famously said "all models are wrong, some are useful", but what he forgot to add was that usefulness doesn't just depend on the model.
A model is useful *only with respect to a given target problem*
A model is useful *only with respect to a given target problem*
October 24, 2025 at 8:13 AM
George Box famously said "all models are wrong, some are useful", but what he forgot to add was that usefulness doesn't just depend on the model.
A model is useful *only with respect to a given target problem*
A model is useful *only with respect to a given target problem*