Chetan
@chetan_
Followers
260
Following
58
Media
20
Statuses
368
cofounder @Ultraroboticsco. building robots in nyc
nyc
Joined August 2019
just found out today that our best teleoperator is in med school training to be a neurosurgeon turns out that the best teleoperators are surgeons and gamers 🤯
0
2
8
Artificial Narrow Intelligence will deploy way before embodied AGI, and still be useful Great thread explaining why
Robotics has an unwritten rule. Hardware first. Deployment later. A small team in Brooklyn ignored that rule and still shipped a working robot in months. Here is what they did differently 🧵:
2
10
113
based attitude
@VukRosic99 It is the result of the collective efforts of all the authors; my individual contribution is insignificant.
0
0
0
More robots. More better.
0
0
13
RIP to K-Scale. Really sad to see this outcome. The worst part about hardware companies is that the capex can result in you dying even if you have a product people want If anyone on the team is looking for what's next, we are hiring for some select roles @Ultraroboticsco
0
0
3
price point for the 1x launch seems insane at first glance, but i think it's actually pretty smart they basically want to give the robots away for free so they can collect diverse data. the price is so that the user has some skin in the game and incentive to actually use it
0
0
6
I woke up excited to test my new AI and I'm going to sleep depressed that I spent the whole day debugging why my robot controller won't boot I love robotics!!🫠
3
0
5
I'm surprised that people are surprised by this. Getting low tracking latency with VR controllers is actually super easy. The hard part is getting low video latency from the cameras into the headset.
1
0
6
the fingers on this hand are fucking massive lol. similar to the leap hand in that respect
Boston Dynamics engineers talk about Atlas’ 3-finger hand. 7-DOF, features a palm camera and tactile-sensing fingertips. They chose not to add more fingers for the initial phases of hand design for higher reliability and speed of development.
1
0
1
pov: humanoid robot appears in your sauna lol
0
0
0
this is something that i think is quite underrated -- learning the whole distribution of outputs rather than just the mean can be really helpful (especially for modeling multimodal distributions)
many classic ML problems that could be framed as continuous regression are actually solved better as classification over discretized bins
0
0
1