FOLD! FOLD! FOLD!
An overnight timelapse of one of our live deployments teleoperated from over 7000 miles away. Our robots fold 24/7/365 with no human intervention, handling long-tail edge cases with no downtime.
@BartronPolygon@VilleKuosmanen they look alright. then i have another question, are you using the leader joint angles to train ACT? i know your leader is your iphone so i’m wondering how you’re passing the data
Me and @arbwes are open-sourcing the code to train the low cost robot (<250$) designed by @alexkoch_ai.
This is the policy trained on a sorting task with less than 30 demonstrations.
https://t.co/OzY9CMm0LL
@enjoychang@rosikand@arbwes@alexkoch_ai probably, in the original implementation of ACT they do something called “temporal aggregation” which is essentially averaging multiple model predictions but it’s too slow to run on cpu. would be curious to test it
i'm throwing the first ever AI simulated party.
it's 3 days long.
day 1 and day 2 are in the simulation.
day 3 you pull up irl to Mission Control in sf.
here's how it works:
1. every guest gets an AI character.
2. you customize it to your personality.
3. your character is thrown into a virtual world where it meets everyone else attending the party.
4. the day of the irl party, you get a report of the top 3 ppl to meet and more importantly, who to avoid lmao.
this is the future of irl parties.
drop a 🎉 now and ill send u an invite.
@rosikand@arbwes@alexkoch_ai inference was done on cpu of an M3 macbook pro. also tried inference on ‘mps’ but it does some weird stuff under the hood and the policy didn’t work well anymore