Lightnews — Scholar-powered news

Reposted by Blake Richards

Shahab Bakhtiari @shahabbakht.bsky.social · 2d

Regardless of what explainability/mech interp in AI is actually after, and whether or not they know what they’re searching for, we can confidently say they’re pursuing what systems neuroscience has pursued for decades, with very similar puzzles and confusions.

Mel Andrews @bayesianboy.bsky.social · 2d

What problem is explainability/interpretability research trying to solve in ML, and do you have a favorite paper articulating what that problem is?

2 8 45

Reposted by Blake Richards

Wu Tsai Institute | Yale University @wutsaiyale.bsky.social · 2d

WTI's Inspiring Speaker Series continues this week with Timothy Lillicrap, “Model-based reinforcement language for reasoning in games and beyond”

10.9.25 | 10 – 11:15a
100 College St, Workshop 1116
🔗 wti.yale.edu/event/2025-10/inspiring-speaker-tim-lillicrap

All Yale community members are welcome.

1 2 7

Reposted by Blake Richards

Martin Schrimpf @mschrimpf.bsky.social · 2d

A glimpse at what #NeuroAI brain models might enable: a topographic vision model predicts stimulation patterns that steer complex object recognition behavior in primates. This could be a key 'software' component for visual prosthetic hardware 🧠🤖🧪

1 9 40

Reposted by Blake Richards

Brokoslaw Laschowski @drlaschowski.bsky.social · 3d

Imagine a brain decoding algorithm that could generalize across different subjects and tasks. Today, we’re one step closer to achieving that vision.

Introducing the flagship paper of our brain decoding program: www.biorxiv.org/content/10.1...
#neuroAI #compneuro @utoronto.ca @uhn.ca

3 14 61

Reposted by Blake Richards

Aidan Horner @aidanhorner.bsky.social · 4d

If you're interested in the cognitive neuroscience of memory feel free to email me!

I do experimental psychology, brain imaging (fMRI and MEG) and a bit of modelling. Lab is doing stuff on forgetting, aging, schemas, and event boundaries, but we're not limited to that.

#psychscisky #neuroskyence

Aidan Horner @aidanhorner.bsky.social · 4d

It's that time of year when many start thinking about applying for PhDs. If you're applying for a UK PhD position, here is a blog post I wrote a while back that might be helpful

#cognition #psychscisky #neuroskyence #psychjobs

How to get PhD funding in the UK

It is that time of year again. The leaves are turning golden, red, and orange (or just brown), the nights are drawing in, and there is a chi...

aidanhorner.blogspot.com

3 26 46

Reposted by Blake Richards

Ching-Lung Hsu @hiallen72.bsky.social · 5d

New function for excitatory input to the basals of the CA1
www.pnas.org/doi/10.1073/...

Hippocampal CA2 to CA1: A metaplastic switch for memory encoding | PNAS

The hippocampus is essential for spatial and episodic memory, subserved by CA1 neurons. Hippocampal area CA2, which processes social memory, also m...

www.pnas.org

3 7

Reposted by Blake Richards

Shahab Bakhtiari @shahabbakht.bsky.social · 4d

Interesting paper suggesting a mechanism for why in-context learning happens in LLMs.

They show that LLMs implicitly apply an internal low-rank weight update adjusted by the context. It’s cheap (due to the low-rank) but effective for adapting the model’s behavior.

#MLSky

arxiv.org/abs/2507.16003

Learning without training: The implicit dynamics of in-context learning

One of the most striking features of Large Language Models (LLM) is their ability to learn in context. Namely at inference time an LLM is able to learn new patterns without any additional weight updat...

arxiv.org

1 19 60

Reposted by Blake Richards

Shahab Bakhtiari @shahabbakht.bsky.social · 4d

The results challenge the texture-bias hypothesis of Geirhos et al. (2019).

This is one of those cases where a deep, careful review can add real value.

arxiv.org/abs/2509.20234

🧠🤖 #MLSky

ImageNet-trained CNNs are not biased towards texture: Revisiting feature reliance through controlled suppression

The hypothesis that Convolutional Neural Networks (CNNs) are inherently texture-biased has shaped much of the discourse on feature use in deep learning. We revisit this hypothesis by examining limitat...

arxiv.org

1 6 18

Reposted by Blake Richards

Sarah Dorner @sarahdorner.bsky.social · 7d

Le Conseil d'Outremont (avec de nouvelles marionnettistes) donne leur appui au groupe « Outremont 0 km » qui dénonce les 0 km d'infrastructures de transport actif implantées depuis le mandat d'Ensemble Montréal à Outremont.

Les marionnettes du Conseil d'Outremont donnent leur appui au groupe « Outrémont 0 km ».

2 7

Reposted by Blake Richards

Shahab Bakhtiari @shahabbakht.bsky.social · 10d

So excited to see this preprint released from the lab into the wild.

Charlotte has developed a theory for how learning curriculum influences learning generalization.
Our theory makes straightforward neural predictions that can be tested in future experiments. (1/4)

🧠🤖 🧠📈 #MLSky

Charlotte Volk @charlottevolk.bsky.social · 10d

🚨 New preprint alert!

🧠🤖
We propose a theory of how learning curriculum affects generalization through neural population dimensionality. Learning curriculum is a determining factor of neural dimensionality - where you start from determines where you end up.
🧠📈

A 🧵:

tinyurl.com/yr8tawj3

The curriculum effect in visual learning: the role of readout dimensionality

Generalization of visual perceptual learning (VPL) to unseen conditions varies across tasks. Previous work suggests that training curriculum may be integral to generalization, yet a theoretical explan...

tinyurl.com

1 6 31

Reposted by Blake Richards

Charlotte Volk @charlottevolk.bsky.social · 10d

🚨 New preprint alert!

🧠🤖
We propose a theory of how learning curriculum affects generalization through neural population dimensionality. Learning curriculum is a determining factor of neural dimensionality - where you start from determines where you end up.
🧠📈

A 🧵:

tinyurl.com/yr8tawj3

The curriculum effect in visual learning: the role of readout dimensionality

Generalization of visual perceptual learning (VPL) to unseen conditions varies across tasks. Previous work suggests that training curriculum may be integral to generalization, yet a theoretical explan...

tinyurl.com

1 22 69

Reposted by Blake Richards

Karl Jacoby @karl-jacoby.bsky.social · 14d

Only an administration intent on committing war crimes in the present and future would stoop to calling Wounded Knee a "battle" rather than what it truly was: a massacre of over 250 Lakotas, mainly women, children, and the elderly. 1/

1.1K 7.5K 21K

Blake Richards @tyrellturing.bsky.social · 14d

Who's to say filling in requires activity in V1?

1 6

Reposted by Blake Richards

Martin Jaggi @mjaggi.bsky.social · 15d

We're hiring again for AI research engineering roles: Join the team behind the Apertus LLM, if you share our passion to work on impactful AI that's truly open.

careers.epfl.ch/job/Lausanne...

AI Research Engineers - Swiss AI Initiative

careers.epfl.ch

2 4 5

Reposted by Blake Richards

Kris Jensen @kristorpjensen.bsky.social · 16d

I’m super excited to finally put my recent work with @behrenstimb.bsky.social on bioRxiv, where we develop a new mechanistic theory of how PFC structures adaptive behaviour using attractor dynamics in space and time!

www.biorxiv.org/content/10.1...

9 85 210

Blake Richards @tyrellturing.bsky.social · 17d

Strong recommend!!!

Really fascinating exploration on the links between life, intelligence, prediction, and computation.

(Disclosure: @blaiseaguera.bsky.social is now my boss.)

The Transmitter @thetransmitter.bsky.social · 17d

In his new book, published today, artificial-intelligence researcher @blaiseaguera.bsky.social draws from computer science, machine learning and neuroscience to provide a nuanced examination of the fundamental aspects of intelligence.

#neuroskyence

www.thetransmitter.org/learning/wha...

‘What Is Intelligence?’: An excerpt

In his new book, published today, Blaise Agüera y Arcas examines the fundamental aspects of intelligence in biological and artificial systems. In this excerpt from Chapter 4…

www.thetransmitter.org

1 1 15

Blake Richards @tyrellturing.bsky.social · 17d

A bit, depends on your definition... Our team at Google is doing neuro-inspired research, but the goal is not neuro insights, per se. (Though we will do a bit of that.)

6

Reposted by Blake Richards

Alexander Doria @dorialexander.bsky.social · 17d

Might not be the consensual opinion on here but totally true. More exciting developments right now than at any time in the past.

5 4 70

Blake Richards @tyrellturing.bsky.social · 17d

Probably less... you're welcome. 😝

1 6

Blake Richards @tyrellturing.bsky.social · 17d

4/4) Keep your eyes out for what our Paradigms of Intelligence team will be producing in the coming months and years. I’m pumped about the work and I’m confident that this group will produce some major breakthroughs in the near future to make AI more efficient and robust. 🙂 🧠 🤖

3 14

Blake Richards @tyrellturing.bsky.social · 17d

3/4) I’m going to maintain a reduced position at @mcgill.ca and @mila-quebec.bsky.social, so don’t consider me as having completely abandoned academia. (I'm lucky to be where I am...) But, I’m keen to get more time to work on some bigger frontier problems I couldn’t tackle in my own lab.

1 15

Blake Richards @tyrellturing.bsky.social · 17d

2/4) This is a big step for me, having spent my adult life in academia. But, there was no way I could pass up an opportunity to work with some of the smartest iconoclasts in the business, including @blaiseaguera.bsky.social himself, @dileeplearning.bsky.social, and many others.

1 14

Blake Richards @tyrellturing.bsky.social · 17d

1/4) I’m excited to announce that I have joined the Paradigms of Intelligence team at Google (github.com/paradigms-of...)! Our team, led by @blaiseaguera.bsky.social, is bringing forward the next stage of AI by pushing on some of the assumptions that underpin current ML.

#MLSky #AI #neuroscience

Paradigms of Intelligence Team

Advance our understanding of how intelligence evolves to develop new technologies for the benefit of humanity and other sentient life - Paradigms of Intelligence Team

github.com

23 11 180

Reposted by Blake Richards

Qingqing Zhang @qqzhang.bsky.social · 18d

Does predictive coding work in SPACE or in TIME? Most neuroscientists assume TIME, i.e. neurons predict their future sensory inputs. We show that in visual cortex predictive coding actually works across SPACE, just like the original Rao+Ballard theory #neuroscience
www.biorxiv.org/cgi/content/...

4 26 85

Reposted by Blake Richards

Julie Charlton @juliecharlton.bsky.social · 18d

🛑 Attention researchers who train freely-moving rodents!

Jess Breda and I developed a protocol for training center-port nose fixation 61% faster than a previous curriculum while keeping violation rates low.

Preprint here: www.biorxiv.org/content/10.1...

FixGrower: An efficient and robust curriculum for shaping fixation behavior in rodents

Center-port fixation is a common prerequisite for many freely-moving rodent tasks in neuroscience and psychology. However, typical protocols for shaping this behavior are non-standardized and ineffici...

www.biorxiv.org

1 9 35