I find it annoying that people who for many years have been at the forefront of "Scale, baby, scale!" are now lecturing on the importance of actually thinking. There is a community of people who have been thinking all along. Without expending terawatt-hours of electricity for no good reason

This gets to the heart of a big problem in neuroscience. Since we don't have a formal model of neural computation and since everything is important to people who study it, we are trapped in this "but my thing matters too!" mode Prioritization of what is important is necessary, even if harsh

I can't stop reading Konrad's thread here, and it just drives something home that I'm increasingly appreciating: Neuroscientists really really really REALLY want the brain to be low dimensional. So much so that the low dimensionality is often assumed, and as Konrad gets to here, that causes problems

Konrad is doing an amazing take-down here, and at least on this point I fully agree. The "brain is analog" argument is a terrible strawman that holds back neuro. Analog dynamics may help, but the theoretical scaling of neurons mostly come from event-driven parallelism. arxiv.org/abs/2507.17886

Our paper on a neuromorphic algorithm for solving finite element models is out, which is exciting for a number of reasons. But really I want to talk a bit about how this relates to understanding the brain and #NeuroAI. 🤖🧪🧠🧵 1/ www.nature.com/articles/s42...

Nothing like flying out of a city when all the SfN folks are flying in. Satellite meetings are where it's at. 3 over 3 days was a little intense. But smaller meetings are more fun. Enjoy SfN everyone!

I was amused to find myself quoted, kind of, in the last paragraph of this; don't know if I've ever been the punchline quote before!

This is a great thread, and I think it hits on one of the biggest challenges in neuroscience that I hope NeuroAI can impact. Until we can rigorously define what computations are occurring in the brain, we can't make any real progress in our functional understanding. We need constraints. 🧠🤖🧪

We put the FlyWire connectome on the Loihi 2 neuromorphic platform 🤖🧠🧪🪰 arxiv.org/abs/2508.16792

An interesting comment and discussion that gets to the deeper question of whether NeuroAI is a move towards something new or repackaging of old tired approaches with a shiny AI paint job.

Traffic in Seattle is really bad...

I'm a big fan of computing with neurons, stem cells, organoids, and energy efficient AI. But we really have to stop coming up with systems (bio or inorganic) that are a few thousand neurons and claiming wins. AI models have BILLIONS of neurons. GPUs are quite efficient for the models they run 🤖🧠🧪

Spiking neural networks people, this message is for you! The annual SNUFA workshop is now open for abstract submission (deadline Sept 26) and (free) registration. This year's speakers include Elisabetta Chicca, Jason Eshraghian, Tomoki Fukai, Chengcheng Huang, and... you? snufa.net/2025/ 🤖🧠🧪

For a few years I have said that neuromorphic is specialized general purpose, like GPUs, but with different advantages. In this preprint I try to put some substance to that claim. There are real theoretical advantages, but they aren't obvious. 🧪🧠🤖 www.arxiv.org/abs/2507.17886

This is an interesting thread getting to th heart of the neuroscience / AI disconnect. I think we need better "comp neuro for dummies" options, but there is a widely held view among engineers and computing people that neuroscientists obsess over details for no reason.

Dear editors: If you want me to review a paper, give me a form with two entries: comments for authors, optional comments for editor. Maybe (maybe) ask for accept/revise/reject recommendation, but that is really your job. Please don't ask me to answer 12!!! separate questions... 🧪

Lately my feed has been full of stories like "The brain's learning is more complex than Hebb thought!" and "Does the adult brain have new neurons?". Hebb was almost 100 years ago and the neurogenesis "debate" is 30 years old. We are worse than Hollywood in terms of rehashing the same old stories 🧠🧪

Bill Dally presenting the rise in necessary compute for AI I see why this is good for Nvidia. But why would this be considered good for anyone else?

At the @cra-hq.bsky.social CCC Computing Futures Symposium this week - exciting (and interesting) times to be talking about the future of computing research! As part of this event, last night we demo'd neuromorphic and its potential impact on ModSim and #NeuroAI at the US Senate!

With the demise of Twitter, I greatly miss the often critical but honest takes on new results. LinkedIn's AI community is all hype and often wrong. I would like to see Bsky prioritize honesty and truth over becoming an echo chamber, particularly with #NeuroAI new results that risk being misused 🧠🧪🤖

This RealID chaos is so strange. I am quite certain I have had a compliant ID for about 10, maybe even 15 years. What is going on that some states are so backwards that they can't figure this out? And I used to live in some pretty dysfunctional states that could even mange this.

I'm unsure whether I agree with this. On the one hand, questions matter the most. On the other hand, I'm not convinced we (as neuroscientists) know the right questions to ask.

I'm not a fan of either, but at least with "ANNs to describe the brain" the idea is new. For 50+ years physicists have been trying to shoehorn the brain into mean field approaches with little justification beyond "it would be so convenient" If you don't want to think about the brain, don't study it

All right world, I think it is time to write...

Has anyone looked into using LLMs to interpret the information content of words in different languages? For instance, German has many words for 'the' (thus presumably higher information content), and Spanish verbs capture contextual uncertainty much differently than English

My only comment about NIH indirects is that universities should pay postdocs and grad students more (a direct cost), and maybe-just maybe-seriously looking at where the money is going may eventually end up directing more $ to those who actually do the research.

If the endless "some people optimized LLMs" news is starting to bore you, check out our preprint on neuromorphic-compatible neural circuits that can solve sparse linear systems. arxiv.org/abs/2501.10526 This opens up some exciting new doors for brain-like algorithms! 🧪🧠🤖 #NeuroAI #Neuromorphic

I've never been directly funded by NIH, but in honor of this amazing organization with great people, here is my talk from last fall when they hosted me. One of the best visits I have had anywhere. oir.nih.gov/wals/2024-20...

I'm very excited that this review on scalable neuromorphic hardware, headed up by Dhireesha Kudithipudi, has finally come out in Nature🧪🧠🤖 rdcu.be/d7atq Neuromorphic hardware is ready for prime time. There are things to keep advancing, but I think the bigger opportunities lie in how we use it!

I've been watching with amusement the kerfuffle of Nvidia's CEO saying quantum computing is 15+ years away. Setting aside all of the self-interest of the leading GPU company bashing an emerging technology, it is illustrative to think about what it means for neuromorphic technologies and NeuroAI. 🧪🧠🧵

Nothing like journals pre-emptively guilting about reviews due on December 26th...

People may enjoy this 🧵 about why scale matters for adult neurogenesis. An oldie paper that probably has meaning for machine learning and #NeuroAI today. 🧪🧠💻

We have a News and Views in Nature Computational Science with a back of envelope estimate on how expensive whole brain neural simulations can be. 🧠🧪 Long story short... very very expensive This maybe ends up behind a pay wall (sorry ...), but check it out! www.nature.com/articles/s43...

Has anyone applied the Stanford Marshmallow experiment to academic disciplines? In particular it seems highly relevant re. the penchant of comp sci and EE for conference papers with shallow results over more impactful journal papers that take more work but clearly have greater staying power. 🧪🧠

It is kind of amazing the scale and speed of the original neural simulations. We have come a long way from 386 processors and 8 MB of RAM ~35 years ago (Wilson and Bower 88) My PhD work ~20 years ago was 20x more neurons, albeit much simpler, and many more time steps. Today we can sim billions 🧠🧪

One of my worries about #NeuroAI is that if it gets defined too broadly the AI bros will just steal it as a marketing term for their near-term optimizations of commercial apps. NeuroAI requires the brain. Being lazy and ignoring the brain just because the brain is hard should be out of scope.

My new office art

I don't know what the next wave of AI will be, but I strongly suspect it won't be another "please ignore the nuclear power plant behind the curtain" boondoggle for GPU companies. Efficiency matters. The future of AI is the brain. 🧠🧪 #NeuroAI

So cool! But I think neural plasticity has this even this dynamic range beat!

Any last minute submissions for neuromorphic computing, spiking algorithms, or neural theory? Send them over to NICE 2025, we will be in Heidelberg in March. Deadline in two days! niceworkshop.org/nice-2025/

All fields need to do this periodically so that their value proposition can be clearly articulated to the people who provide the funding. Agreeing on goals and the key questions to answer is not the same as prescribing solutions (which is where I think past large neuro efforts tripped up).

Also, while I'm talking about getting to know neuromorphic... As in the discussion on #NeuroAI, there are a lot of definitions and views of what #neuromorphic is. Not all of it is what neuroscientists would consider brain-inspired (just like not all AI is related to the brain).

Wondering what neuromorphic computing is? 🧠💻 To peak into the brain-inspired crowd in neuromorphic, I recommend looking at the NICE Conference. We've been debating the intersection of neuroscience, AI, and computing for over a decade now, and many past talks are online niceworkshop.org/nice-2025/

Seriously... AI researchers complaining about lack of GPUs is like car companies complaining about domestic oil refinery capacity. Maybe take a hint?

This often gets cited as a problem. And it is for labs that wish to use AI while focusing on experiments and other questions. But for those studying AI, shouldn't this be a clear sign of what the problem is? Maybe study alternatives that aren't overkill? #NeuroAI 🧠🧪 www.nature.com/articles/d41...

A perspective from the #NeuroAI #BRAIN meeting last week, which was awesome and I’m honored to have been a part of.🧠🧪 One thing is bugging me though. I continue to think that is a trap to look at where ANNs are in alignment with the brain. I think we have to look at the problem from the other way

#NeuroAI workshop today: great discussion on how we will recognize success? My answer: when we can understand major neurological / mental health disorders well enough for cures Brains suffer from diseases of computation. Today we don't understand the computation so we don't know how to fix them

Crazy thought at the #NeuroAI meeting. How would our knowledge / views / theories of the brain be different if we were first taught that serotonin (and it's wild cadre of receptors) was the principle neurotransmitter of the brain, and glutamate and GABA were just fast fine tuning modulators? 🧠🧪

This is a very exciting workshop to be part of, and everyone should tune in. Advancing neural theory with computing as a guide is not just the future of BRAIN but should be part of the future of neuroscience research in general!