Tool use in insects: Assassin bugs apply resin to their forelegs before a stingless bee hunt. This makes the bees attack the bug in just the right position to be caught! Videos will worth watching www.pnas.org/doi/full/10....

Ariel Krakowski interviews me on his podcast about brains, AI, plasticity, connectomics, consciousness, and everything in between. open.spotify.com/episode/4m33...

On today’s science headlines : A newly discovered species of carnivorous caterpillar on the Hawaiian island of Oahu has earned the nickname “bone collector” for its eerie habit of adorning itself with the remains of its prey, such as ant heads and fly wings. 🪲 🧪

Emergence and maintenance of modularity in neural networks with Hebbian and anti-Hebbian inhibitory STDP with @raphaelbergoin.bsky.social and @gzamora-lopez.bsky.social finnally published in @plos.org Computational Biology #neuroskyence #compneurosky journals.plos.org/ploscompbiol...

Join us in Jerusalem for a great program!

High-Dimensional Dynamics in Low-Dimensional Networks. New preprint with a former undergrad, Yue Wan. I'm not totally sure how to talk about these results. They're counterintuitive on the surface, seem somewhat obvious in hindsight, but then there's more to them when you dig deeper.

Our work elaborating on, simplifying, and extending earlier formulations of the dynamics of calcium-based plasticity has been published in JOCN. Bottom line is: 'calcium tells the synaptic weight where it's going and how fast it gets there.' link.springer.com/article/10.1...

How does a neuron get its activity? 👀 Check out our latest preprint, where we tracked the activity of the same neurons throughout early postnatal development: www.biorxiv.org/content/10.1... see 🧵 (1/?)

Anyone here do EM/connectomics? Any pointers to why I don't really see differences in luminance symmetry between E and I syns (defined acc. to presynaptic cell type?)

Adventures in proofreading: Journal has online system to edit proofs. Work for an hour, am auto-logged out. Log back in, all edits are gone. Try editing on PDF instead. For a substantial number of edits, it's impossible to highlight the relevant text because UNCORRECTED PROOF gets in the way.

Now out in PLOS CB! We propose a simple, perceptron-like neuron model, the calcitron, that has four sources of [Ca2+]...We demonstrate that by modulating the plasticity thresholds and calcium influx from each calcium source, we can reproduce a wide range of learning and plasticity protocols.

[T]he priesthoods are a perfect environment for memetic plagues...Those plagues that successfully capture them have found some way to thread this balance, basing themselves in overarching social theories outside the specialties’ competence to assess.

Sharing our most recent paper, in which we find that electrical synapses in C. elegans drive action selection by "filtering" sensory information. C. elegans can learn temp, move across the temp. gradient and then track that temp. How? Configuration electrical synapses www.cell.com/cell/fulltex... 1/

Is our brain infinitely flexible or constrained? Oby & colleagues cleverly uses BCIs to test what cortical activity can/can't be generated quickly. Highly recommend! www.nature.com/articles/s41... And you can get a tl;dr + my takes on why this is exciting here: www.nature.com/articles/s41...

More people should leave academia to start companies. There's a lot of overlap in the skill sets. If you can turn an idea into a published paper, there's a decent chance you can also turn an idea into a product. Probably more impactful in the long run than writing papers that 5 people will read.

Ever wondered what different layer-5 PC types do for learning? Our work suggests that one (IT PCs) does representational learning whereas the other (ET PCs) encodes representational value! A great exp-theory collaboration with the Larkum and Takahashi's labs!

Long shot, but does anyone have a version of the Allen Microns connectomics dataset with *all* synapses labeled as E or I (based on synapse morphology, not presynaptic neuron identity?)

Excited to see @thetransmitter.bsky.social feature our preprint on how the neural code changes across the cortical hierarchy! A multi-region perspective on categorical selectivity 🧱 and geometric dimensionality 📐 Blueprint thread coming soon— after I am done with panettone digestion 🥮!

Key-value memory is an important concept in modern machine learning (e.g., transformers). Ila Fiete, Kazuki Irie, and I have written a paper showing how key-value memory provides a way of thinking about memory organization in the brain: arxiv.org/abs/2501.02950

What's a good review of information flow in the cortex that discusses cell types, layers, circuits, and inhibition?

Dissecting origins of wiring specificity in dense cortical connectomes Looks hot! Generative modeling for neural networks -- trying to get insight into the wiring rules of dense connectomes. www.biorxiv.org/content/10.1...

Excited to announce the publication of our team's work suggesting a new framework for understanding hashtag#synapse hashtag#development and diversity in the mammalian hashtag#brain in Nature Neuroscience rdcu.be/d2100 🧵

Son hormigas resolviendo un problema geométrico y es para flipar en colores.

DAY 22: Advent of Comp Neuro 🎄🤖🧠🧪 Memories in a network can drift at the single neuron level, but persist at the population level! 🍎🍏 “Drifting assemblies for persistent memory: Neuron transitions and unsupervised compensation” www.pnas.org/doi/full/10....

People climb the career ladder to get a management position (PI) and then once they get there they find that they have less time for non management-related things <surprised Pikachu face>.

🚨 New Paper! Can neuroscience localizers uncover brain-like functional specializations in LLMs? 🧠🤖 Yes! We analyzed 18 LLMs and found units mirroring the brain's language, theory of mind, and multiple demand networks! w/ @gretatuckute.bsky.social, @abosselut.bsky.social, @mschrimpf.bsky.social 🧵👇

Every cog neuro talk: “Here’s a picture of downtown Tokyo. How do humans navigate this rich visual environment? Let’s find out. In experiment 1, three rats looked at triangles while I measured one neuron from each…”

Neuronal dynamics can be #balanced without the need of strong external currents, as usually done, here with A. Politi we explain how (open access article 2024) #complexity #neuroscience #neuroskyence #compneurosky pubs.aip.org/aip/cha/arti...

DAY 15: Advent of Comp Neuro 🎄🤖🧠🧪 Heterogeneity of neural properties can improve some computational functions, but homogeneity also has its advantages! It's complicated 😅 “Neural heterogeneity controls computations in spiking neural networks” @rgast.bsky.social et al www.pnas.org/doi/epub/10....

Reposting new preprint investigating sex-differentially expressed (DE) genes in human ventromedial hypothalamus and arcuate 🧠. Data identifies extensive sex DE, which is linked to genetic risk for sex-biased disorders including autism, depression, and schizophrenia www.biorxiv.org/content/10.1...

This paper on the usefulness of diversity of neuron types is pretty cool: journals.plos.org/ploscompbiol...

1/ Okay, one thing that has been revealed to me from the replies to this is that many people don't know (or refuse to recognize) the following fact: The unts in ANN are actually not a terrible approximation of how real neurons work! A tiny 🧵. 🧠📈 #NeuroAI #MLSky

Wonderful work by Eckert et al. on biochemical models of habituation in single cells: www.sciencedirect.com/science/arti...

This slide is making the rounds, I figured I should weigh in on it because my PhD thesis was basically about the premise of the syllogism. Long story short, it's not obviously false.

Unlocking the 'black box': Scientists reveal AI's hidden thoughts #DL #AI #ML #DeepLearning #ArtificialIntelligence #MachineLearning #ComputerVision #AutonomousVehicles #Robotics #LLM #VLM #LVLM https://buff.ly/4gicgXB

Check out our new preprint! We show there is a better learning algorithm for #compneuro than gradient descent (GD): exponentiated gradients (EG). tl;dr: EG respects Dale's law, produces weight distributions that match biology, and outperforms GD in biologically relevant scenarios. 🧠📈 🧪

Thing I've noticed when I use AI to code (Cursor with Claude) - there is a very pronounced 80/20 effect, where the AI is very good at reducing the work needed to get 80% of the job done - the basic architecture, the boilerplate, but is much worse at the final 20% - cosmetics, QA, etc.

DAY 9: Advent of Comp Neuro 🎄🤖🧠🧪 The shape and speed of neural trajectories, as wells as manifold orientation are shaped by the network structure and inputs, which we can control! 💫 "Dynamic control of neural manifolds" www.biorxiv.org/content/10.1...

New paper out in #Neuron: A general theory of sequential working memory in prefrontal cortex and RNN/SSMs with their exact neural mechanism. Plus unifying this new mechanism with the alternate mechanism of hippocampal cognitive maps! (1/9) www.cell.com/neuron/fullt...

Another fascinating study from the Yasuda lab: www.nature.com/articles/s41... CaMKII inhibition immediately after training impairs memory transiently. Memory recovers a day later!

DAY 5: Advent of Comp Neuro 🎄🤖🧠🧪 Heavy-tailed distributions enable smooth transitions to chaos and other biologically realistic activity. “Edge of Chaos and Avalanches in Neural Networks with Heavy-Tailed Synaptic Weight Distribution” journals.aps.org/prl/abstract...

Why are journals so adamant about not accepting revised manuscripts with trackchanges? There's literally one button you press to show/hide the revisions, it's clearly the easiest way for reviewers to see what revisions you've made. Not allowing it is the exact opposite of what makes sense.

From Randy Bruno's group, learning a CS+/- task leads to the apical tufts in S1 distinguishing the two stimuli: elifesciences.org/reviewed-pre... It would be amazing if we could know what the downstream relationship to behaviour is of the cells, to determine if this is for credit assignment. 🧠📈 🧪

I wonder how many scientific followers I can successfully alienate by metalposting. Anyway, what a way to open a song.

🧪 E11 Bio is excited to share a major step towards brain mapping at 100x lower cost, making whole-brain connectomics at human & mouse scale feasible (🧠→🔬→💻). Critical for curing brain disorders, building human-like AI systems, and even simulating human brains. Read more: e11.bio/news/roadmap

Our new study is out on BioRXiv! tinyurl.com/pfcmap We (@carlenlab.bsky.social lab) mapped the mouse PFC using single-unit activity! We recorded ~23,000 neurons across cortical/subcortical regions and profiled spont. firing patterns to reveal what separates the PFC from other brain regions. 1/n

Questionably popular opinion: for most intents and purposes, extant SOTA LLMs *are* AGI. They are clearly smarter than the average person about nearly every conceivable subject, and they are often superior to experts. The main thing they lack is embodiment and adeptness at physical tasks.

Rant: If someone is doing independent scientific work in a lab, even if they're a PhD student (or an undergrad), they are not a 'trainee'. You are being 'trained' when you are being instructed on the basics of how to do the work, not when you are a productive contributing member of a lab.