Aidan

@aidan_sirbu

PhD @ Mila and McGill | Paradigms of Intelligence Team @ Google | Deep learning, Recursive Reasoning, Embodied Intelligence, Computational Neuroscience

Montreal, QC

aidansirbu.com

Joined June 2023

95 Following

145 Followers

Replies Reposts Media

Pinned
Aidan
@aidan_sirbu
Oct 21
1/14) Virtual animals are a long-standing goal in neuro-AI, but they should actually behave like real animals. We present SCAMPER: a virtual rat for RL studies that moves naturally during both exploration, and goal-directed behavior. Project page: aidansirbu.com/SCAMPER-site/
5
Aidan
@aidan_sirbu
1h
Blake has been an incredible advisor to me and many other students and postdocs! I’ve met very few people so well versed in both neuroscience and machine learning. If you’re looking for a postdoc, this is a no-brainer…
Blake Richards
@tyrell_turing
1h
1/4) We're hiring in my lab, **please share**! 🚀 There are two postdoc opportunities in cog neuro and AI. This is for a new project supported by a Schmidt Polymath Award. This project aims to build an engaging video game to collect data for neurofoundation models. 🧠🤖
Aidan
@aidan_sirbu
Oct 1
Check out our latest work! We show that Neural Cellular Automata can solve complex visual reasoning tasks using fully local communication. This brings us closer to AI models that naturally map to decentralized hardware, potentially enabling power-efficient mesh communication.
Mayalen Etcheverry
@mayalen_etc
Oct 1
Can complex multi-step reasoning emerge purely from cells that only talk to immediate neighbors? Happy to share our paper “Reasoning with Neural Cellular Automata (NCA)”, from our team at Google, Paradigms of Intelligence 🧵👇
GIF
1
Aidan
@aidan_sirbu
Sep 11
Great to see more work in this direction!
Ayhan Suleymanzade
@ayhozade
Sep 11
Can neural networks learn to think by denoising? Excited to share Thinking with Looped Flows! We train recurrent reasoning with local denoising objectives. It is a new approach for reasoning that achieves SoTA on ARC-AGI-1 and 2 among looped models. In collaboration with
Aidan
@aidan_sirbu
Sep 3
Fresh off the press! We show that you can augment a denoising model with a hidden memory state and remove time-step conditioning to turn it into an anytime solver. More work to come soon...
Mariia Drozdova
@MariiaD_ML
Sep 3
1/7) What happens when we add a persistent memory state and remove timestep conditioning from a denoising model? We obtain an anytime iterative solver! With less than 250k-params, we obtain a 99.9% accuracy on Sudoku-Extreme and 98.3% on Maze-Unique. arxiv.org/abs/2609.01449
GIF