Tessera
Folgen

Pouya Bashivan 🇮🇷🇨🇦

@bashivan.bsky.social

Husband, dad, computational neuroscientist, @mcgillu.bsky.social and Mila

56 Posts248 Follower122 Folgt
Pouya Bashivan 🇮🇷🇨🇦@bashivan.bsky.social · last mo.

Do you have any fun memories from past CCN conferences? Please share your experiences and any photos you have from previous CCNs! We are organizing an event for #CCN2026 to remember previous years.

CogCompNeuro@cogcompneuro.bsky.social · 2mo ago

CCN began at Columbia in 2017. As #CCN2026 returns to New York City, a special Back2NY event will feature a crowd-sourced panel discussion on how the community has grown, prior challenges, and future perspectives. Share your experiences in the community survey: 2026.ccneuro.org/back2ny/

Pouya Bashivan 🇮🇷🇨🇦@bashivan.bsky.social · 5mo ago

Our work on building world models from episodic memories is accepted to #ICLR26! Led by (then undergraduate, now masters student) @herbiehe.bsky.social Check out the tweeprint!

Herbie(Zizhan) He@herbiehe.bsky.social · 5mo ago

New paper 🚨 #ICLR26 Most world models predict the future from a past trajectory. But neuroscience suggests that such inference can instead be made from temporally independent experiences. We built the Episodic Spatial World Model (ESWM), a model that does exactly this: Video abstract [1/2]

Pouya Bashivan 🇮🇷🇨🇦@bashivan.bsky.social · 5mo ago

Super proud of this work co-led by graduate students Ali Saheb Pasand and @johanoosterman.bsky.social and fantastic collaboration with Aaron Courville and @pcastr.bsky.social

Pablo Samuel Castro@pcastr.bsky.social · 5mo ago

New paper 🚨 "Stable Deep Reinforcement Learning via Isotropic Gaussian Representations" Deep RL suffers from unstable training, representation collapse, and neuron dormancy. We show that a simple geometric insight, isotropic Gaussian representations, can fix this. Here's how 👇

Pouya Bashivan 🇮🇷🇨🇦@bashivan.bsky.social · 6mo ago

Very proud of this work by @motaharehpr.bsky.social ! Glad to see it accepted as a talk at #Cosyne2026 Tldr; training a neural net to search for objects in natural scenes made it not only behave like humans but also made it converge on similar computations and representations as primates!

Motahareh@motaharehpr.bsky.social · 6mo ago

Excited to give a talk at #Cosyne2026 about my PhD work! We show that RNNs trained on visual search converge on brain-like solutions, producing primate-like behavior and neural representations. Happy to chat if you're at Cosyne! 📅 March 15, 2026 📍 Lisbon, Portugal www.biorxiv.org/content/10.1...

Yoshua Bengio@yoshuabengio.bsky.social · 7mo ago

Today we’re releasing the International AI Safety Report 2026: the most comprehensive evidence-based assessment of AI capabilities, emerging risks, and safety measures to date. 🧵 (1/19)

Taylor Webb@taylorwwebb.bsky.social · 8mo ago

Very much looking forward to this series of workshops on the computational ingredients of reasoning. We have an amazing lineup of speakers from diverse backgrounds, and there will be lots of opportunities for discussion. Please consider attending!

IVADO@ivado.bsky.social · 8mo ago

🧠 Computational Ingredients of Reasoning: participate in the next #IVADO Thematic Semester, which will be held in Montreal from January to March 2026. ➡️ Register now: event.fourwaves.com/fr/thematics... @glajoie.bsky.social @taylorwwebb.bsky.social @lampinen.bsky.social

Pouya Bashivan 🇮🇷🇨🇦@bashivan.bsky.social · 8mo ago

Interesting essay by Tim Dettmers. Although I don’t fully agree with all the predictions, the contrast between China and North America’s approach to practical AI will potentially be defining in years to come

hardmaru@hardmaru.bsky.social · 8mo ago

“Why AGI Will Not Happen” by Tim Dettmers. timdettmers.com/2025/12/10/w... This essay is worth reading. Discusses diminishing returns (and risks) of scaling. The contrast between West and East: “Winner takes all” approach of building the biggest thing vs a long-term focus on practicality.

Why Scaling Is Not Enough I believe in scaling laws and I believe scaling will improve performance, and models like Gemini are clearly good models. The problem with scaling is this: for linear improvements, we previously had exponential growth as GPUs which canceled out the exponential resource requirements of scaling. This is no longer true. In other words, previously we invested roughly linear costs to get linear payoff, but now it has turned to exponential costs. Frontier AI Versus Economic Diffusion The US and China follow two different approaches to AI. The US follows the idea that there will be one winner who takes it all – the one that builds superintelligence wins. Even coming short of superintelligence of AGI, if you have the best model, almost all people will use your model and not the competition’s model. The idea is: develop the biggest, badest model and people will come. China’s philosophy is different. They believe model capabilities do not matter as much...
Pouya Bashivan 🇮🇷🇨🇦@bashivan.bsky.social · 9mo ago

Attending the Montreal AI and Neuroscience (MAIN) Conference this week? #MontAIN2025 We have put together some exciting educational workshops on cognitive benchmarking large models, RL and video games and dynamical systems! More info and registration here: main-educational.github.io/program/

Program - MAIN educational 2025

Website of the educational workshop organized during the Montreal Artificial Intelligence and Neuroscience Conference 2025

main-educational.github.io

Pouya Bashivan 🇮🇷🇨🇦@bashivan.bsky.social · 9mo ago

There’s a lot of debate about “superhuman AI” that might end it for us. All this made me think if that AI, presumably conscious-like, would build even better AI superseding itself? Specially if, it has been trained on these debates. If it choses not to, then it might become the smartest AI ever (?)

Blake Richards@tyrellturing.bsky.social · 9mo ago

PSA for academics involved in designing admissions systems: Setting a specific time-of-day deadline is *ri*dic*ul*ous* and super annoying!!!! Do you really care if I submit this lettter at 6PM rather than 4PM? Were you planning on reviewing my letter that evening? Get real... #academia

Pouya Bashivan 🇮🇷🇨🇦@bashivan.bsky.social · 9mo ago

VLMs are truly caught in the middle! We found that they are great at describing what they see AND great at reasoning from text. But, they fail to connect them without an explicit text bridge. Come talk to us about this and more at #Neurips2025

Zihan@zhweng.bsky.social · 9mo ago

1/9 🚨Thrilled to share "Caption This, Reason That", a #NeurIPS2025 Spotlight! 🔦 Meet us at #2112, 3 Dec 11 a.m. We analyze VLM limitations through the lens of Cognitive Science (Perception, Attention, Memory) and propose a simple "Self-Captioning" method that boosts spatial reasoning by ~18%. 🧵👇

Pouya Bashivan 🇮🇷🇨🇦@bashivan.bsky.social · 9mo ago

I still see much focus on single factor coding in studies of medial temporal cortex. Although our results in primate hippocampus indicates a more mixed code www.biorxiv.org/content/10.1...

Spatially and non-spatially tuned hippocampal neurons are linear perceptual and nonlinear memory encoders

The hippocampus has long been regarded as a neural map of physical space, with its neurons categorized as spatially or non-spatially tuned according to their response selectivity. However, growing evi...

biorxiv.org

Earl K. Miller@earlkmiller.bsky.social · 9mo ago

I remember when we used to think that all neurons were specialized. Mixed selectivity: when neurons stopped looking like specialists www.nature.com/articles/s41... #neuroscience

Pouya Bashivan 🇮🇷🇨🇦@bashivan.bsky.social · 10mo ago

The future of Canadian research

a setwin service bus with people getting on it

ALT: a setwin service bus with people getting on it

media.tenor.com

Shahab Bakhtiari@shahabbakht.bsky.social · 10mo ago

Can’t agree more. Still trying to wrap my head around the budget news. I want to move on, but honestly can’t. Yes, there’s some self-interest here, but adding another well-established researcher to an already deprived funding landscape isn’t exactly encouraging.

Blake Richards@tyrellturing.bsky.social · 10mo ago

Great Blueprint from @arnaghosh.bsky.social on our newest paper on representational geometry! tl;dr: we find that during pretraining LLMs undergo consistent cycles of expansion/recuction in the dimensionality of their representations & these cycles correlate with the emergence of new capabilities.

Arna Ghosh@arnaghosh.bsky.social · 10mo ago

LLMs are trained to compress data by mapping sequences to high-dim representations! How does the complexity of this mapping change across LLM training? How does it relate to the model’s capabilities? 🤔 Announcing our #NeurIPS2025 📄 that dives into this. 🧵below #AIResearch #MachineLearning #LLM

New paper titled "Tracing the Representation Geometry of Language Models from Pretraining to Post-training" by Melody Z Li, Kumar K Agrawal, Arna Ghosh, Komal K Teru, Adam Santoro, Guillaume Lajoie, Blake A Richards.