Jon Miller Schwartz
@JonMSchwartz
Building and deploying practical, general-purpose robots @ultraroboticsco
If you’re not jumping onto the table at dinner yelling with excitement about your startup… change something
The future will be filled with an amazing menagerie of robots. I can’t wait
This guy built an auto aiming trash can for his messy coworkers
And it's only getting better from here...
Can you tell what’s autonomous and what’s teleop? Neither can our customers. This video is 100% playback speed, and 50% autonomous.
Robot TV is the best TV
Checking out what the robot friends are up to in the warehouses whilst policy training
A little bit of joyful design goes a long way in making your robot approachable and not look like terminator
#dinov2 image classifier amazes me: after just 2,000 examples trained on L40 in 10 min it's able to label HOURs of robot footage @Ultraroboticsco
Are humanoid robotic hands really necessary? Are they better than simpler end effectors? Will all robots eventually converge on them? At @UltraRoboticsCo, we’ve had the rare opportunity to test a range of end effectors in real customer deployments, manipulating diverse items…
Nothing beats real robots that produce real data
I wrote a fun little article about all the ways to dodge the need for real-world robot data. I think it has a cute title. sergeylevine.substack.com/p/sporks-of-agi
Misha and his team are world class in a way all startups dream of being. Huge launch.
Engineers spend 70% of their time understanding code, not writing it. That’s why we built Asimov at @reflection_ai. The best-in-class code research agent, built for teams and organizations.
Is it me, or does it feel like everyone is trying to get their piece before the singularity hits?
Note that at 28:46 a bag falls off the conveyor. What happens to that order? Until robots can act with as much agency and ability as humans, we’ll need specialized software/hardware-based workflows to reliably account for edge cases.
Watch Helix's neural network do 60 minutes of uninterrupted logistics work Helix now incorporates touch and short-term memory and it's performance continuously improves over time
Both sides of the humanoid argument are right: we should and will build specialized equipment and robots, AND humanoids will be massively valuable for filling in the gaps between and at the tail ends
Anyone have an extra Trossen ViperX 300 S you want to sell? I'm buying!