Mohamad H. Danesh
@mo_danesh
Followers
233
Following
4K
Media
58
Statuses
533
CS PhD @McGillU and @mila_quebec, working on 🍒 and 🤖 stuff / ex- @LetsUnifyAI, @NUSComputing, @EngineeringOSU
Montréal, Québec
Joined February 2017
Work led by Stanley Wu, with yours truly, Simon Li, Hanna Yurchyk, Amin Abyaneh, @aahoussaini, David Meger, and Hsiu-Chin Lin from @mcgillu and @Mila_Quebec. Gratefully supported by Google DeepMind, FRQNT, and NSERC.
0
0
2
4/ VOCALoco: - Boosts robustness & safety over Deep RL benchmarks - Deploys zero-shot on real ANYmal-D - Predicts terrain difficulty before failure - Scales to new skills (stairs, obstacles, gaps) by plugging in new low-level policies!
1
0
1
3/ Given a terrain heightmap, two neural modules predict for each skill: 1️⃣ Viability: can this skill safely traverse the terrain? 2️⃣ Cost-of-Transport: how energy-efficient will it be? We then pick the safest, most efficient skill, all in real time! https://t.co/B37RSckogl
1
0
1
2/ Instead of relying on a single end-to-end RL policy, VOCALoco selects between multiple specialized locomotion skills (eg, walking, stair ascent, descent) depending on the terrain beneath the robot.
1
0
1
🧵1/ Our paper "VOCALoco: Viability-Optimized Cost-aware Adaptive Locomotion" has been accepted to RA-L! 🎉 We introduce a modular locomotion framework that makes quadrupeds safer, more energy-efficient, and adaptive across terrains: https://t.co/Fmdv8L5wIb
arxiv.org
Recent advancements in legged robot locomotion have facilitated traversal over increasingly complex terrains. Despite this progress, many existing approaches rely on end-to-end deep reinforcement...
1
1
3
Exciting news! We're thrilled to announce the appointment of Professor @hugo_larochelle as Mila's new Scientific Director! A deep learning pioneer and former head of Google's AI lab in Montreal, Hugo's leadership will be pivotal in advancing AI for the benefit of all. Read the
12
28
257
(1/3) Today’s VLAs: pure System 1. The “System 1 + System 2” pitch (Pi0, Helix, survey) flatters them with psychology they don’t have. Great for investors who read Kahneman, less so for accuracy.
Great intro to vision-language models by @chris_j_paxton Vision-Language-Action Models and the Search for a Generalist Robot Policy
1
4
59
i just realized PCA is Fourier analysis without the sines and cosines, but with your data’s own modes.
55
86
2K
new conference, new review process, same old chaos. ICML was a disaster, NeurIPS is keeping the vibe, and we'll see about ICLR. why improve the process at all when you can just remix randomly
This year's ICML review process was… something. The acknowledgement button felt like a placebo, and reviewer engagement was basically a ghost town, a solid 0/10. Not exactly a goldmine of constructive feedback.
0
1
5
That's exactly the kind of advisor you do not want to do a PhD with!
At this point, you can use any AIs (perplexity, chatgpt, etc) as a PhD advisor on whatever topic you're getting deep into. It's pretty good. I have my preference. The core point: Research advice is no longer an elite academic university thing.
10
11
260
I’ve found the laziest AC or maybe they found me. Look people: if you get invited to be an AC at a top conference like… I don’t know, ICML, and then go M.I.A. during the entire process until you have ChatGPT write your decision summary, and then you don’t even bother to check
4
9
190
Introducing π-0.5! The model works out of the box in completely new environments. Here the robot cleans new kitchens & bedrooms. 🤖 Detailed paper + videos in more than 10 unseen rooms: https://t.co/64ze7ToQup A short thread 🧵
15
101
702
This year's ICML review process was… something. The acknowledgement button felt like a placebo, and reviewer engagement was basically a ghost town, a solid 0/10. Not exactly a goldmine of constructive feedback.
1
0
24
RL is so back! (well, for some of us, it never really left) https://t.co/1XPx29bGPj
awards.acm.org
For developing the conceptual and algorithmic foundations of reinforcement learning.
6
31
320
🚨 Montreal Robotics Summer School 2025 - Applications Open! 🚨 📅 When: August 10-15, 2025 📍 Where: Mila – Quebec AI Institute (In-person) 💡 What: 1 day of activity + 5 days of intense robotics & AI training (theory + hands-on!)
3
16
95
Today, we are open sourcing the Pi0 model including checkpoints, training code and more
github.com
Contribute to Physical-Intelligence/openpi development by creating an account on GitHub.
7
28
298
I work in robotics. I find things uniquely stressful rn like never before. Everyday can be a "GPT moment in robotics", especially for AI related ones. I know colleagues have mental illnesses due to the stress in the robotics domain, despite good lab atmosphere and nice PI.
Do you work in AI? Do you find things uniquely stressful right now, like never before? Haver you ever suffered from a mental illness? Read my personal experience of those challenges here: https://t.co/OFZmSizl6f
19
27
618
📢 I am on the JOB market this year 📢 I am looking for both faculty and research scientist positions. My research makes AI agents useful and safe for humans. I enable them to effectively convey uncertainty, ask for help, learn from human feedback, and pursue goals that benefit
3
18
43
Unitree's secret weapon lies in the scaling laws of their robotic hardware. Regardless of how robust a policy may seem, the reality for many robotics labs is that a significant amount of time is spent repairing robots. I am genuinely curious—how many B2-W units have been broken
Unitree B2-W Talent Awakening! 🥳 One year after mass production kicked off, Unitree’s B2-W Industrial Wheel has been upgraded with more exciting capabilities. Please always use robots safely and friendly. #Unitree #Quadruped #Robotdog #Parkour #EmbodiedAI #IndustrialRobot
0
4
30