1/3 Making Spot 🐕🤖by Boston Dynamics ‼️ aware of the people around him to avoid collisions🙋🙋‍♀️: The Harper #IROS2024 benchmark (intelligolabs.github.io/HARPER/) considers 6 onboard cameras (the robot PoV🥽) and an OptiTrack system (the panoptic PoV 🎦) as ground truth data...

The Boston Dynamics demo from the side.

# RLHF is just barely RL Reinforcement Learning from Human Feedback (RLHF) is the third (and last) major stage of training an LLM, after pretraining and supervised finetuning (SFT). My rant on RLHF is that it is just barely RL, in a way that I think is not too widely

Meet our AI-powered robot that’s ready to play table tennis. 🤖🏓 It’s the first agent to achieve amateur human level performance in this sport. Here’s how it works. 🧵

When I train an RL agent to play soccer

Il nuovo corso di laurea in Ingegneria dell'Informazione 100% Università di Verona! Ti interessa programmare robot e sistemi intelligenti? Qui tutte le informazioni: corsi.univr.it/?ent=cs&id=118…

Amazing timelapse of a SpaceX rocket booster returning to Earth from space.

Why aren't nut allergies properly handled by many airlines? There are special meals for everything—low salt, meals for religious diets, low-calorie, vegan, vegetarian, even NO SPICE meals—but no reliable nut-free option. I've faced this issue several times recently. Anyone else?

🫣😉"THIS POST IS ERORS WE WHILE COMMUNICATING" 😜🧐... and the fact that AI machines 🤖have to cope with that! Today at #IROS2024 Francesco Taioli gave a very impactful oral, showing that in Language-driven Navigation, with errors, performance drops.Curious?Here: intelligolabs.github.io/R2RIE-CE/

double-blind peer review

Thanks again for prize and the follow-up promotion of our work! Francesco Taioli should be named for all his hard work 😋

People ask me what's next. The GOAT points the way. Physical AI. Embodied Agents. Robotics. That's what's next.

When your date sends you the simplest message

New Scaling for Humanoid Robotics Coding your way of this won't work, we simply need a step-change in capabilities to scale to a billion-unit robot level To be successful in dynamic environments such as the home, robots must to able to generalize on-demand

"But one thing I do know for sure - there's no AGI without touching, feeling, and being embodied in the messy world"

📢The Human-aware Embodied AI (HEAI) workshop #IROS25 is live and accepting submissions! 📜Check out the CfP and our amazing invited speakers lineup in the flyer below! If you're working on these topics, we'd love to see your work. 🔗 More info: heai-iros25-workshop.github.io

Vancouver ❤️

Saw this in a paper today. This made my day. openreview.net/pdf?id=duGygkA…