Jascha Achterberg

Jascha Achterberg

@achterbrain.bsky.social

Neuroscience & AI at University of Oxford and University of Cambridge | Principles of efficient computations + learning in brains, AI, and silicon 🧠 NeuroAI | Gates Cambridge Scholar www.jachterberg.com

2.2K followers920 following181 posts

Jascha Achterberg reposted

Nicolas Skatchkovsky
@nskat.bsky.social

We're organising a #Cosyne workshop on biologically-inspired AI on Monday 16th with @achterbrain.bsky.social! 🧠 We've got an incredibly exciting array of speakers, and you have the possibility to sign up to present your poster on #NeuroAI! More info below ⬇️ #compneuro #neuroscience #neuroskyence

Jascha Achterberg
@achterbrain.bsky.social

Going to Cosyne? There will be a "Biologically-inspired Artificial Intelligence: Challenges and Opportunities" workshop on Monday 16th! 🧠 Exciting lineup of speakers alongside the opportunity to present your #NeuroAI poster, more info on program & poster signup below! #compneuro #neuroscience

View full thread
Jascha Achterberg
@achterbrain.bsky.social

We are also hosting a NeuroAI poster session during the lunch break! If you are presenting a NeuroAI-related poster at Cosyne, you are welcome to put it up in our workshop room so attendees can find relevant work they might otherwise have missed. Signup: forms.gle/kzefm2FtVuRk...

forms.glePresent your poster at the Cosyne NeuroAI workshop on 16th March 2026On the 16th March there will be a NeuroAI-specific workshop at Cosyne ("Biologically-inspired Artificial Intelligence: Challenges and Opportunities"; information on: https://sites.google.com/view/cosy...
Jascha Achterberg
@achterbrain.bsky.social

Super looking forward to the workshop! Alongside the talks and poster session, we are also working on an industry careers in NeuroAI meet-up for Early Career Researchers. All details and updates can be found on our website: sites.google.com/view/cosyne-...

sites.google.comCosyne NeuroAIArtificial intelligence has become a transformative force in scientific research and society at large, yet current models face significant limitations. Transformer-based large language models, despite...
Jascha Achterberg
@achterbrain.bsky.social

Going to Cosyne? There will be a "Biologically-inspired Artificial Intelligence: Challenges and Opportunities" workshop on Monday 16th! 🧠 Exciting lineup of speakers alongside the opportunity to present your #NeuroAI poster, more info on program & poster signup below! #compneuro #neuroscience

Jascha Achterberg
@achterbrain.bsky.social

Enjoyed @giacomoi.bsky.social commentary in PNAS on how #NeuroAI and Neuromorphic Engineering should come together to allow brain circuit motifs to positively influence the design of computing systems: Biological fidelity: The engine driving the neuromorphic renaissance www.pnas.org/doi/full/10....

pnas.orgPNASProceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...
Jascha Achterberg
@achterbrain.bsky.social

Giacomo's commentary was in response to this great recent paper by Iqbal et al., also in PNAS: "Biologically grounded neocortex computational primitives implemented on neuromorphic hardware improve vision transformer performance" www.pnas.org/doi/10.1073/...

pnas.orgPNASProceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...

Jascha Achterberg reposted

Adam J. Eisen
@adamjeisen.bsky.social

How do brain areas control each other? 🧠🎛️ ✨In our NeurIPS 2025 Spotlight paper, we introduce a data-driven framework to answer this question using deep learning, nonlinear control, and differential geometry.🧵⬇️

Jascha Achterberg reposted

Rui Ponte Costa
@somnirons.bsky.social

A beautiful summary of our paper! Thank you @neurosock.bsky.social

neurosock@neurosock.bsky.social

The hippocampus is not a library, it is a simulation engine. HPC is known for storing maps of the environment but not so known for generating planned trajectories. This paper proposes that recurrence in CA3 is crucial for planning. A🧵with my toy model and notes: #neuroskyence #compneuro #NeuroAI

Jascha Achterberg reposted

Cian O'Donnell
@cianodonnell.bsky.social

One example of how multiplexing might be implemented in the brain was shown in the great work by Thomas Akam with Dmitri Kullmann, a decade ago. The papers are cited well but the general multiplexing idea never really took the field by storm as much as it deserved www.nature.com/articles/nrn...

nature.comOscillatory multiplexing of population codes for selective communication in the mammalian brain - Nature Reviews NeuroscienceThe function of brain oscillations remains unclear, although a role in controlling the flow of signals among anatomically connected networks has been proposed. In this Opinion article, Akam and Kullma...
Cian O'Donnell@cianodonnell.bsky.social

100%. I'm a fan of the idea of multiplexing here, that different receivers can decode different information from the same broadcasted signal.

Jascha Achterberg reposted

Dan Goodman
@neural-reckoning.org

Postdoc fellowship opportunity for ECRs (<3 yrs post-PhD). Note that if you want to apply to work with me as your mentor, our dept has an internal deadline of Dec 4th so please email me asap. Our internal process is shorter than the full application. 🤖🧠🧪 royalcommission1851.org/fellowships/...

royalcommission1851.orgResearch FellowshipsOur Mission to “increase the means of industrial education and extend the influence of science and art upon productive industry” Supplemental Charter of 1851
View full thread
Jascha Achterberg
@achterbrain.bsky.social

We also find that while complex regions are needed to learn complex tasks, these tasks are eventually moved toward simpler regions, similar to how you may struggle the first time when learning a new skill, but slowly get better with practice.

Jascha Achterberg
@achterbrain.bsky.social

This new model opens a whole new world of analysing multi region interaction across trials and tasks! More analysis and findings can be found in our paper linked below. Work lead by Jack Cook, and with great help from @danakarca.bsky.social and @somnirons.bsky.social ! arxiv.org/abs/2506.02813

arxiv.orgBrain-Like Processing Pathways Form in Models With Heterogeneous ExpertsExamples of such pathways can be found in the interactions between cortical and subcortical networks during learning, or in sub-networks specializing for task characteristics such as difficulty or mod...
View full thread
Jascha Achterberg
@achterbrain.bsky.social

With these three features in place, we find that our third criterion of distinct pathways is also met. While baseline models exhibit largely random expert usage patterns, our models exhibit highly structured pathways between regions that reliably emerge during learning.

Jascha Achterberg
@achterbrain.bsky.social

Furthermore, we find that these pathways mirror our expected behavior of pathways in the brain! We find that difficult tasks need to be learned in more complex regions, similar to how you need to think “harder” when learning how to solve a difficult math problem.

View full thread
Jascha Achterberg
@achterbrain.bsky.social

Our third contribution is expert dropout. Without this feature, we find models suffer large performance deficits when experts outside of the active pathway are disabled. However, we would want models to be primarily dependent on the experts that are most being used.

View full thread
Jascha Achterberg
@achterbrain.bsky.social

We then set three criteria to determine whether pathways had formed: (1) Consistency: Models trained on the same tasks should have similar pathways (2) Self-sufficiency: Pathways should be primarily reliant on their own experts (3) Distinctness: Many distinct pathways should be present

Jascha Achterberg
@achterbrain.bsky.social

We then identify three inductive biases that yield pathways that meet each of these criteria. The first of these is a routing loss that penalizes the use of more complex experts during training, and the second scales this loss by the model’s performance on the task being solved.

Jascha Achterberg
@achterbrain.bsky.social

We first needed to create a model in which we could study pathway formation. We chose a Heterogeneous Mixture-of-Experts architecture, in which information can be dynamically routed to computational experts, or regions, of varying sizes. We train model on 82 tasks of varying complexity (ModCog)!

Jascha Achterberg
@achterbrain.bsky.social

Check out this cool new work lead by @pengfei-sun.bsky.social !

Dan Goodman@neural-reckoning.org

Psst - neuromorphic folks. Did you know that you can solve the SHD dataset with 90% accuracy using only 22 kb of parameter memory by quantising weights and delays? Check out our preprint with @pengfei-sun.bsky.social and @danakarca.bsky.social, or read the TLDR below. 👇🤖🧠🧪 arxiv.org/abs/2510.27434

arxiv.orgExploiting heterogeneous delays for efficient computation in low-bit neural networksNeural networks rely on learning synaptic weights. However, this overlooks other neural parameters that can also be learned and may be utilized by the brain. One such parameter is the delay: the brain...

Jascha Achterberg reposted

Pengfei
@pengfei-sun.bsky.social

With my great advisors and colleagues, @achterbrain.bsky.social @zhe @danakarca.bsky.social @neural-reckoning.org, we show that if heterogeneous axonal delays (imprecise) can capture the essential temporal structure of a task, spiking networks do not need precise synaptic weights to perform well.

Dan Goodman@neural-reckoning.org

Psst - neuromorphic folks. Did you know that you can solve the SHD dataset with 90% accuracy using only 22 kb of parameter memory by quantising weights and delays? Check out our preprint with @pengfei-sun.bsky.social and @danakarca.bsky.social, or read the TLDR below. 👇🤖🧠🧪 arxiv.org/abs/2510.27434

arxiv.orgExploiting heterogeneous delays for efficient computation in low-bit neural networksNeural networks rely on learning synaptic weights. However, this overlooks other neural parameters that can also be learned and may be utilized by the brain. One such parameter is the delay: the brain...

Jascha Achterberg reposted

Martin Schrimpf
@mschrimpf.bsky.social

Come be our colleague at EPFL! Several open calls for positions 🧪🧠🤖 * Neuroscience www.epfl.ch/about/workin... (deadline Oct 1) * Life Science Engineering www.epfl.ch/about/workin... * CS general call www.epfl.ch/about/workin... * Learning Sciences www.epfl.ch/about/workin...

epfl.chFaculty Position in NeuroscienceThe School of Life Sciences at EPFL invites applications for a Tenure Track Assistant Professor position in Neuroscience. At EPFL researchers develop and apply innovative technologies to understand br...

Jascha Achterberg reposted

Steven Scholte
@neurosteven.bsky.social

#CCN2025 is over. Over 5 days there were 6 fantastic keynotes, 550 posters, 3 community events, 3 keynote & tutorials, 3 generative adversarial collaborations, 8 Satellite events, 1 community lunch meeting, 1 cross-conference hackathon, 1 competition, coffee all day, stroopwafels on day 1,

Jascha Achterberg reposted

The Viking (Gunnar Blohm)
@gunnarblohm.bsky.social

I'm curious if any senior nonhuman primate Neuro-AI researchers would be interested in joining Queen's University if we were to obtain a research chair position (full professor level)? Could you please send me a confidential message to indicate your interest? Gunnar.blohm@queensu.ca

Jascha Achterberg reposted

CogCompNeuro
@cogcompneuro.bsky.social

The registration desk for the first day is now open! If you are looking for information to get to the venue, you can find it here: 2025.ccneuro.org/venue-inform... Look out for the CCN banners!

CogCompNeuro@cogcompneuro.bsky.social

Welcome to the first day of #CCN2025 ! 🧠 Today, we're starting with networking, opening remarks, and two fantastic keynotes from Nancy Kanwisher and Anna Schapiro. More details in the thread below 👇

Jascha Achterberg reposted

CogCompNeuro
@cogcompneuro.bsky.social

Less than 2 weeks until CCN 2025 in Amsterdam! Here's everything you need to know to prepare for the 8th Cognitive Computational Neuroscience conference, August 12-15 at University of Amsterdam 🧠

Jascha Achterberg reposted

Dan Goodman
@neural-reckoning.org

I wrote an article earlier in the week arguing that we need to give junior researchers more independence earlier, and this should be our focus, not moonshot mega projects led by senior researchers. I was surprised how much agreement I'm seeing. So next question: how do we do this?

The Transmitter @thetransmitter.bsky.social

Large-scale projects run the risk of stifling scientific independence. Instead, let’s explore alternative mechanisms of collaboration, writes @neuralreckoning.bsky.social. #neuroskyence www.thetransmitter.org/funding/neur...

thetransmitter.orgNeuroscience needs to empower early-career researchers, not fund moon shotsLarge-scale projects run the risk of stifling scientific independence. Instead, let’s explore alternative mechanisms of collaboration.

Jascha Achterberg reposted

Dan Levenstein
@dlevenstein.bsky.social

#NeuroAI finds itself facing an interesting question these days: 1) which of these now-many schemes for bio-plausible credit assignment actually operate in the brain? -and if more than one- 2) how the hell do they operate when they, inevitably, interact?

Friedemann Zenke@fzenke.bsky.social

1/6 Why does the brain maintain such precise excitatory-inhibitory balance? Our new preprint explores a provocative idea: Small, targeted deviations from this balance may serve a purpose: to encode local error signals for learning. www.biorxiv.org/content/10.1... led by @jrbch.bsky.social