Jason Ma, cofounder of Dyna Robotics, discusses building reliable robot foundation models for real-world manipulation at the Humanoids Summit.
Ma says lab results need to be tested against real deployment conditions. Customer work can reveal useful tasks, supply better data, and show researchers which capabilities still need improvement.
Dyna chose restaurant napkin folding to test whether a general vision-language-action model could handle a demanding task consistently. In a 24-hour run, its model succeeded 99.4% of the time, including when napkins became tangled or lighting changed. A restaurant deployment was folding about 1,000 napkins per day.
After a conventional training approach reached about 80% success, Dyna added a reward model that scores video frames during execution. Its progress estimates help identify where a run fails, so the team can collect focused corrections and train the model to recover from mistakes.
Ma describes deploying a T-shirt-folding model at a robotics conference in Korea without gathering new data there. It handled a new robot, table, lighting, background activity, and shirts it had not encountered before, completing the full process from basket to folded stack.
Dyna reports learning celery chopping and cup stacking with under an hour of training data, despite neither task appearing in pretraining. For longer sequences such as making breakfast or setting a dinner table, Ma says the models reached robust performance with under five hours of data.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.
Jason Ma, Dyna Robotics co-founder and chief scientist, joins Latent Space on site to discuss Dyna-2, robot learning, hardware, and commercial deployments.
Jason Ma, co-founder and AI research lead at Dyna Robotics, presents Dyna 2’s approach to robot learning and commercial deployment at the Fellows Forum on Day 2.
Jason Ma, a robotics researcher at the University of Pennsylvania who worked on the paper at Google DeepMind, joins RoboPapers hosts Michael and Chris to discuss using vision-language models to assess robot task progress and data quality.
Jensen Huang, CEO of Nvidia, describes how AI is being used inside the company and predicts that robotics will reshape factories; AI Explained host discusses these claims with robotics researcher Jason Ma and IPPR report author Carsten Young.
Jason Ma, researcher at the University of Pennsylvania, presents research on safer and reward-free offline reinforcement learning at a GRASP Lab seminar.
