6 Comments
User's avatar
minerva's avatar

I’m skeptical, but wish them luck. What they’re doing isn’t unique. If it works, anyone else can quickly copy them. Also self driving is still struggling with a data only end to end approach if Tesla deployments are anything to go by.

Chris Paxton's avatar

Data is a moat; vertical integration is a moat. it might work!

Hugo's avatar

Nice data collection and business strategy.

I was also wondering if there is a transfer gap. A human hand in a glove and a robot gripper have different contact dynamics, different compliance, and different force profiles. Do you have any information about that?

Their robots are so cute, by the way.

Avik De's avatar

It’s not totally clear from the act-1 blog post I just skimmed but I assume the $400 data collection glove records some kind of fingertip pressure or other tactile data, in addition to maybe some kind of state estimate (maybe encoders or imu).

Viktor Chak's avatar

Question: Will there be more edge cases of in-home robot or self-driving car?

-R.Fumes's avatar

I have to say, I dont get the in-home robot craze beyond being primed on Jetsons cartoons and having a robo-slave fetish. Robots in the work place have a different set of concerns and are generally purpose built, look at how robots work in an amazon fulfillment factory vs teleoperating a home robot on subscription

The biggest question is how to reliably interpret sensory data. I’m not as worried about a robot cracking my glassware while pouring me a glass of water, I’m worried about how soon onto stepping on a toddlers foot will the bot readjust