So this happened an hour ago. I don’t think I’ve seen red this deep in a long time. You could see it quite clearly with the naked eye. #aurora

Super interesting. For training small LLMs, they get results comparable to those trained on an order of magnitude more tokens by replacing normal pre-training (which ingests huge amounts of human-written text) with exclusively synthetic text derived in a structured way from Wikipedia.

I love machine learning

“Our soldiers fought for the shared values of freedom and democratic rights for all. However, these soldiers returned from the war and quickly realized those freedoms and rights did not equally apply to them as they did their non-native comrades.” #IndigenousVeteransDay Lest We Forget

Discover the inspiring journey of Maude Lizaire, the first Afro-descendant woman to earn a PhD at Mila, and a strong advocate for more inclusive science. mila.quebec/en/news/maud...

A new paper with Bogdan Georgiev, Javier Gomez-Serrano, and Adam Zsolt Wagner: "Mathematical exploration and discovery at scale" arxiv.org/abs/2511.02864. Further discussion is at terrytao.wordpress.com/2025/11/05/m...

“Edge of chaos” dynamics are long recognized as a computationally potent dynamical regime that avoids vanishing gradients during learning and allows greater memory and expressivity of a system. This stark difference surprised us, and we think it can help explain our results on neural adaptation.

Indeed, Lyapunov exponents show that fixed points for RL models largely stay near 0, showing these networks’ dynamics lie at the edge of chaos. Whereas SL models’ dynamics are contractive and orderly, keeping very little information in memory for long and having stereotyped expressivity.

Two Metis Nation-Saskatchewan officials were in Ottawa this week seeking funding and urgent collaborative support for culturally appropriate addictions treatment centers. This is in response to increased addictions and gang violence seen in the area.

1/3 Thank you to CIFAR and partners in DSET for bringing me to Banff to speak on my research: The challenge of hidden gifts in multi-agent reinforcement learning arxiv.org/abs/2505.20579. We introduce a novel task on reciprocity with a scarce resource; take what you need, leave what you don’t.

I’ve been finding papers that experiment with ANNs at the hardware level very interesting recently. Such as diffusive memristors www.nature.com/articles/s41...

First Nation of Na-Cho Nyӓk Dun Elder Walter Peter never imagined he’d see himself as a hologram. But thanks to the wonders of technology, Peter was recently turned into a nearly life-size 3D version of himself.

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

I love artificial intelligence

[1/9] Excited to share our new paper "A Pragmatic View of AI Personhood" published today. We feel this topic is timely, and rapidly growing in importance as AI becomes agentic, as AI agents integrate further into the economy, and as more and more users encounter AI.

[9/9] Read the full paper here: arxiv.org/abs/2510.26396 Coauthors: Sasha Vezhnevets, @xtan, @WilCunningham

Writing a beautiful symphony is easier than doing the dishes in a random home

Uber driver: What’s your name? Dan? Me: yes (it’s not) Uber driver: What a unique name

The simplex algorithm is super efficient. 80 years of experience says it runs in linear time. Nobody can explain _why_ it is so fast. We invented a new algorithm analysis framework to find out.

Albania Prime Minister Edi Rama announced AI minister Diella is "pregnant" with 83 AI "children" to assist Socialist Party MPs www.news18.com/amp/world/al...

You don't need to be intelligent to survive! Think of viruses. Even if you think it's purely a next token predictor, when you allow the system to operate autonomously, it can survive simply by virtue of the actions it takes

I love computer science

still true, i reckon

Paper , on social hierarchy and foraging decisions, accepted in Physical Review Research . Should be online soon on this link: doi.org/10.1103/5vfb...

What if we did a single run and declared victory

Haven't seen it posted a lot, so I will repost the link medium.com/@TmlrOrg/tml... Cool stuff, may encourage write the papers in different tempo

The collection has been a source of controversy for the Vatican amid the broader museum debate over the restitution of cultural goods taken from Indigenous peoples during colonial periods.

I really enjoy reading equations in formal fields (e.g math, computer science, et.c) due to all the colours and shapes that appear in your mind but making low math public outreach posters is fun too @ivado.bsky.social @mila-quebec.bsky.social

A nice feeling after long long hours

Louise BigEagle’s documentaries focus on giving her Nakota language a platform, which only has a handful of speakers left in Saskatchewan. Her latest documentary is called Zehaga O'aknbik'ta, which means ‘Our Way Back Home’ in Nakota.

🚀 Excited to share a major update to our “Mixture of Cognitive Reasoners” (MiCRo) paper! We ask: What benefits can we unlock by designing language models whose inner structure mirrors the brain’s functional specialization? More below 🧠👇 cognitive-reasoners.epfl.ch

To follow up on the asymptotic proof that the self-correction term works with any number of agents or coalitions, here are the results for 3 agents. Policy gradient agent's performance suffers with more agents but self correction still stabilizes learning arxiv.org/abs/2505.20579

In 2018, Charles Murray challenged me to a bet: "We will understand IQ genetically—I think most of the picture will have been filled in by 2025—there will still be blanks—but we’ll know basically what’s going on." It's now 2025, and I claim a win. I write about it in The Atlantic.

So that’s what model welfare is for

'This study will examine the impacts of the unstated paternity policy and the gradual decrease of status Indians in some First Nations bands,' says Battiste. www.aptnnews.ca?p=278355

Here’s a music video that I made for Danger Mouse & Black Thought. youtu.be/LmK978jFxMk?...

I am seeing a lot of people reposting Lakota Man today bc it's Indigenous Peoples Day. A reminder that he is not well liked amongst many (most?) Natives on social media. Many of us blocked him a long time ago. Some of the reasons why are in this article. www.dailydot.com/irl/lakotama...

Hard to believe I’m watching snow fall now after a day like yesterday. Beautiful walk full of amazing nature encounters in the oldest migratory bird sanctuary in North America. Saw a few snow geese 😂. I’ll share more after I get through the photos! 🌿 #birds #prairie

Here is my plan to make Bluesky more fun and active:

NEW on our #DeeperLearning blog People balance being kind vs. being honest — and #LLMs should too. New research shows training choices often favor informativeness over kindness, but prompting can induce sycophancy. Read more: bit.ly/3Wqrtxl

1/8 A month ago I wrapped a 4-month project with MuTek Forum’s AI Ecologies Lab led by Sarah Mackenzie: the research arm of Montréal’s 25-year electronic music festival. Why entropy can make AI more resilient Event: ra.co/events/2206981

My eye colour apparently changed after 6 years

We're finally out of stealth: percepta.ai We're a research / engineering team working together in industries like health and logistics to ship ML tools that drastically improve productivity. If you're interested in ML and RL work that matters, come join us 😀

I am on one transformer paper from 3 years ago and ICLR flooded my bids with RLVR & RLHF :S

w/ James Cohan, @jacobeisenstein.bsky.social, and Kristina Toutanova Paper link: arxiv.org/abs/2509.22445

A huge thank you to my collaborators @shahabbakht.bsky.social and Christopher Pack for their guidance on this project. We’d love to hear your thoughts and comments! The preprint: www.biorxiv.org/content/10.1...

9. We hypothesized that the efficacy of the learning curricula depends on how many distinct, useful visual features the brain recruits to solve the task - curricula which lead learners to rely on fewer, more essential visual features will result in better generalization.

5. In this study, we leveraged ANNs to develop a mechanistic predictive theory of learning generalization in humans. Specifically, we wanted to understand the role of **learning curriculum**, and develop a theory of how curriculum affects generalization.

🚨 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

so fucking tiresome to get emails like this whenever I write about residential school history, truly. people who believe that graves don't exist if they can't see the bodies with their own two eyes possess the critical thinking skills of a baby playing peekaboo.