PRO ROBOTS’ host examines Figure’s home robot trial, its performance against rival systems, and the data and safety questions surrounding household robots.
Figure tested its robot in 30 rented San Francisco homes, with people nearby to intervene for safety. It attempted toy cleanup, towel folding, and bed making without training in those specific homes or on the test items. The robot completed 237 of 420 attempts, for a 56% overall success rate.
Bed making scored 67%, towel folding 62%, and toy cleanup 40%; a toy attempt failed if even one item was missed. The scoring allowed imperfect results, including loose folds and misaligned bedding, to count as passes. Sunday Robotics CEO Tony Zhao cited a 99.1% rate for Memo, but that counted individual items rather than full multi-step tasks.
In Figure’s comparison, the model trained with human-action video succeeded in 56% of attempts, versus about 8% without that prior training. Figure Index gathers people’s uploaded household and work videos; the company said uploads exceeded 16 million and that the archive was growing rapidly. Figure also presented results suggesting more training data reduced model error, though the host noted that better predictions on test videos do not necessarily mean better household performance.
The results support a benefit from human video, but the tested chores do not establish dexterity for tasks such as fastening clothing or handling fragile objects. Rodney Brooks questioned whether video can teach touch-dependent hand skills, while Sergey Levine warned that human demonstrations may not show the easiest methods for robots. The segment also discusses disputed claims about Figure’s safety and past deployments, remote human control of 1X NEO, and wheeled home robots such as Sunday’s Memo and Weave’s laundry system.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.
Kalil 4.0 discusses humanoid robots for homes, factories, retail, and companionship, comparing emerging companies and their plans across the United States, China, Europe, and elsewhere.
sudoremove discusses Sunday Robotics’ ACT-2 laundry-folding results and data strategy, then turns to Walden Robotics, founded by Russ Tedrake and former Toyota Research Institute colleagues.
Claire Zau of ZAUEY visits Sunday Robotics with CEO Tony Zhao and head of product Camila to discuss Memo, a household robot that learns chores from human demonstrations.
Tony Xiao, founder of Sunday Robotics, explains to the Sources Podcast hosts how the company is developing Memo, a home robot trained with data from wearable gloves.
Rock of Over The Horizon talks with robotics designers Maharad Farmani and Scott Walter, and hand surgeon Gustav Anderson, about Sunday Robotics’ Mimo and its prospects for home and commercial work.