Machenta exits stealth: one demo, no real-world training data
Machenta says its robot foundation model learns a new task from a single demonstration, trained without a single pixel of real-world data. Meanwhile John Carmack points out no robot alive can change the batteries in a jack-o-lantern.
The big one
Brad Porter says general dexterous manipulation is two to three quarters away
Brad Porter says he made a specific prediction on stage at the MIT AI conference that fluid generalized physical AI for dexterous manipulation is 2 to 3 quarters away, citing his own pre-training recipes, scaling laws and a method for fusing sim and real data.
There is no robot in the world that can change the batteries in a jack-o-lantern.
Shown this week
Machenta exits stealth with $40M and one-demonstration task learning
Machenta came out of stealth with a $40m seed co-led by Creandum and USV, built on a model that Edward Johns says learns a task from one demonstration with no further training.
A model trained entirely in simulation, without a single pixel of real-world data.
FANUC, KUKA and five more put sim-ready robot models on Hugging Face
NVIDIA Robotics published 2,000+ simulation-ready 3D assets free and ungated on Hugging Face, with hardware models from seven OEMs including Comau, FANUC, Franka, Robotiq, Standard Bots, Universal Robots and KUKA. Lukas Ziegler notes these vendors historically kept CAD close and shipped simulation inside their own tools like RobotStudio, RoboGuide and KUKA.Sim.
If your arm isn’t in the library, whatever assembles the workcell picks one that is.