Neurosys

🎓➡️📚🗯💻📊↔️🔬⚙🧠

[Science Magazine @ScienceMagazine]

New in Science: Researchers present LightGen—the first all-optical chip capable of performing challenging advanced generative #AI tasks at speeds and energy efficiencies orders of magnitude beyond today’s traditional electronic hardware.

Learn more: https://t.co/DDPlUPBEqB

[Science Magazine @ScienceMagazine]

Scientists have developed programmable microrobots, each the size of a single-cell organism, which can use onboard systems to self-propel, sense temperatures, and change behaviors autonomously.

Learn more in @SciRobotics: https://t.co/pqkVf6X2Fy

[European Journal of Neuroscience @EJNeuroscience]

Your Brain Has a Learning Shortcut AI Can’t Copy https://t.co/yU7PFOI3om

[Sheila Macrine, Ph.D. @MacrinePhD]

"Beyond "Platonism" lies the khôra. 🏛️This talk explores ontogenesis—the genesis of forms—by moving past physicalism and reified math. Using Simondon & Plato’s actual operators to rethink the encounter between form & matter.
https://t.co/MkvQZqfzq3 @drmichaellevin
#Metaphysics #Plato #Ontogenesis #Philosophy

[Edgar Dobriban @EdgarDobriban]

I wrote a review paper about statistical methods in generative AI; specifically, about using statistical tools along with genAI models for making AI more reliable, for evaluation, etc. See here: https://t.co/0aq8hJqXzo!

I have identified four main areas where statistical thinking can be helpful. These are just a subset of what is out there; other topics have been well-covered in other reviews.

1. Designing "statistical wrappers" around a model, for instance, changing behavior of a trained model (e.g., abstaining), where a score, e.g., an "unsafety score" is too high. The key connection to statistics is to use the quantiles of the loss (on a calibration set) to set the critical threshold, thus enabling conformal-type high probability guarantees.

2. Closely related, methods for uncertainty quantification, which enable the model to express uncertainty in an answer. A crucial component here is "calibration", whereby the uncertainty is required to reflect reality.

3. Statistical methods for AI evaluation: Specifically, tools for statistical inference (e.g., confidence intervals) on model performance. Exciting recent work proposes careful statistical models for leveraging a very small high-quality dataset, possibly combined with much larger low-quality datasets, for accurate evaluation.

4. Experiment design and interventions. Careful AI experiments to understand and steer models may require interventions such as modifying experimental settings in a controlled manner. This brings up connections to classical experimental design in statistics. This connection has largely remained implicit so far, and my review aims to make it more explicit; hoping that experimental design principles will become useful here.

This review references the work of many, including @HamedSHassani @obastani @tatsu_hashimoto @yuekai_sun @CsabaSzepesvari @ml_angelopoulos @stats_stephen @yaniv_romano @yaringal @KilianQW @_onionesque +their teams, and some work that I was also involved in.

Hopefully, my review will be helpful to orient yourself in this exciting area. Nonetheless, since the area is rapidly expanding, it is possible that I missed important references. Please feel free to let me know of anything that I should add/change!

[Danyal @DanAkarca]

Two papers out, on a new paradigm of temporal computation! Our first work funded by @ARIA_research. We're super proud of this, and there's much more coming:

Neural networks, specifically their weights, have become the most useful functional abstraction from the brain. As powerful function approximators they seeded the current era of AI. But there are many more useful abstractions. The brain does so much more than learn weights.

At its core, the brain exploits the structure of the physical world to perform computation. It aligns itself to reality - to time and space. @achterbrain and I have long been working on how to embed neural networks in space and link this to hardware. But there's another dimension: time itself. We think leveraging both is a big part of the puzzle as to why human learning is so efficient.

If this is true, how can we use space and time in neural networks? What would this even mean? We pitched to
@BramhavarSuraj at @ARIA_research one possible way. Several years ago, I stumbled upon theoretical work that made it concrete. Time delays - the physical fact that signals take time to travel - can store memory (and other things, including increasing the number of computable functions). Even in feedforward networks. The delay here isn't overhead as would be traditionally thought but a feature that can be exploited.

TL/DR - our main findings show that you can do computation in neural networks with time, without (much) need for weights. And it's remarkably efficient. We also show it's possible to co-design hardware (we open-source a chip design) with novel architectures that exploit time to maintain long contexts.

In our first paper, we train neural networks to learn delays and weights. The result: state-of-the-art performance on all the temporally complex neuromorphic benchmarks we tested. Crucially, once you're encoding time directly, weights become almost irrelevant. We compress them to 1.58 bits, just positive, negative, or absent (ternary) weights. That's it. Model sizes drop to double-digit kilobytes!

This works because we're finally encoding information the way the task needs it. Time and space. Which just so happens to be... everything in the physical world. Robotics. Embodied systems. Physical intelligence.

In the second paper, we turn to memory. Intelligent systems don't just compute - they hold onto information over long context windows. And doing this efficiently in hardware underlying computation is a hard, unsolved problem.

To solve this, we built a dual memory pathway architecture: fast spiking dynamics plus a compact state-space memory module that evolves much more slowly. Inspired by how the mammalian cortex separates fast somatic spiking from slower dendritic integration. Each layer maintains a tiny amount of working memory - just ~5% of hidden width - that summarises recent activity and feeds back into the network. There are many directions to scale this further. We co-designed the algorithms and hardware together from the ground up. The result: >4× throughput and >5× energy efficiency, beating Intel's Loihi2 and other leading neuromorphic platforms like DenRAM and ReckOn.

We're open-sourcing the chip design so people can build on this.

Incredibly proud of the team: @pengfeisun17, @achterbrain, @neuralreckoning, Zhe Su and @giacomoi

One of our key takeaways: the current AI paradigm is a narrow slice of what's possible. Scaling homogeneous systems only gets you so far. Biological intelligence is deeply heterogeneous: different timescales, different substrates, different specialisations, all co-evolved together. We think the next frontier of scaling means embracing that heterogeneity. Algorithms and hardware aren't separate problems. They need to co-evolve together.

Can't wait to share what we're cooking next.

[Zhiqiang Sha (@zhiqiangsha.bsky.social) @sha_zhiqiang]

Mapping the genetic landscape across 14 psychiatric disorders | Nature
https://t.co/d3Rc7YuEtn

[Nicolas Rouleau @DrNRouleau]

Eggs are a great source of choline, which is a precursor for acetylcholine, one of the most important neurotransmitters for executive cognitive functions.

[Kasra Jalaldoust @causalkasra]

Neuroscientists record two main data modalities: electromagnetic waves and Spiking activities. A longstanding debate in neuroscience is whether the brain waves are essential for understanding how the brain works, or are they merely epiphenomenal. In more scientific terms, do the waves carry a causal influence on the behavior when spiking activities are recorded?

When I first heard about this from @ZabehErfan I assumed neuroscientists are already discussing in causal inference terms; turns out that it is not quite the case. Each neuroscientist has their own qualitative beliefs/knowledge about the brain mechanisms, but to our surprise, they are not using the causal inference semantics to express it.

The debate is not settled despite decades of advancement in recording technologies and statistical modeling breakthroughs, and it is not too surprising from a causal inference perspective: due to Causal Hierarchy Theorem (@eliasbareinboim et al. 2020), no amount of data alone can identify the causal quantities. Thus it is necessary to use a formal language to express the qualitative knowledge, and @yudapearl’s do-calculus is precisely the syntax and semantics we need to settle the spike-wave duality.

In particular, we ground the discussion in Structural Causal Models, and boil down the epiphenomenality claim to assessing a certain inequality:
P(behavior | spikes) ?= P(behavior | spikes, do(waves))

Once the qualitative domain knowledge is properly expressed in causal terms, e.g., a semi-Markovian causal graph, then a graphical criterion for epiphenomenality can be derived. Grounding the discussion in causal language sheds light on many hidden aspects of the problem, even if the debate remains unsettled:

1. What is the inference bottle neck? Is it the amount of data, regimes of data, scope of measurements, or the statistical modeling methods?
2. Data-driven knowledge discovery: what can be said about the causal graph using the data? which forms of heterogeneity and surrogate interventions are useful?

If you’re at @NeurIPSConf come check out our poster along many other cool work at @CauScien workshop!

[Nina Miolane 🦋 @ninamiolane.bsky.social @ninamiolane]

Arrived #NeurIPS2025 ☀️

If you're interested in the interplay of Geometry, Topology, Algebra w/ Neuroscience & AI, I'll give 2 talks on Sunday:

🌐11:30am: The Algebra of Spatial Navigation (Groups & grid cells)

🍩1:30pm: Topological Deep Learning (Complexes & Graphs)

More👇

[Quanta Magazine @QuantaMagazine]

Winter Is Coming: And the Earth’s oldest organisms are prepping. 🥶
https://t.co/BN2IwVRPHq

[Nature Reviews Gastroenterology & Hepatology @NatRevGastroHep]

New online! Mechanisms of chronic abdominal pain in inflammatory bowel disease and implications for treatment https://t.co/xaDJlkFoim

[Nature Computational Science @NatComputSci]

📢Lei Xing and colleagues present an unsupervised DL method that uncovers clear trajectories of brain activity, effectively distinguishing cognitive events, learning stages, and active vs passive movement. https://t.co/mtQEtY9Nwm @lx2015 @StanfordRAdOnc

🔓https://t.co/Nfy7FC8ZH3

[Alessandro Crimi 🧠🧬🔬🩺( @alecrimi.bsky.social ) @Dr_Alex_Crimi]

We have this finally online 🧠🧠🧠:
"Modeling the Spread of Misfolded Proteins in #Alzheimer’s Disease using Higher-Order Simplicial Complex Contagion", big shout to Marcin, @iacopoiacopini,@lordgrilo & V.Latora
https://t.co/7rHR0c2Nm6 Free on MedXiv!
Code:https://t.co/EIBg6GHctS

[Farzaneh Najafi @farznaj]

Our lab’s 1st paper is out on bioRxiv!

What many (including us) thought were temporal prediction/error signals, turn out to be intrinsic interval-timing signals.

Led heroically by my student Yicong @yicong4ever with amazing support from the whole lab.

https://t.co/5bKkcvpC72

[Neurosys @systems_neuro]

"Neural networks have become standard tools in many areas, yet many important statistical questions remain open. This paper studies the question of how much data are needed to train a ReLU feed-forward neural network."
https://t.co/1XrRHFeh0U

[Neuron @NeuroCellPress]

Online now: A posture subspace in the primary motor cortex https://t.co/fIoJbRuMmn

[Tal Linzen @tallinzen]

faculty opening in computational linguistics and cognitive models of language and cognition at the University of Trento (deadline Sep 14): https://t.co/AxTkOU7iFU

I visited this department a couple of times and loved the science, the city, and the proximity to lago di garda!

[Nature Rev Neurosci @NatRevNeurosci]

Rethinking women’s brain health — a Comment article by Julia Sacher & Ingo Bechmann

https://t.co/q15qsfcrSL