Chet Bhateja

Research engineer @DynaRobotics, prev @1x_tech

[Chet Bhateja @ChetBhateja]

Excited to share Dyna-2! Enjoyed collaborating with the team to push the boundaries of cross-embodiment and instruction following. We're just getting started!

[Dyna Robotics @DynaRobotics]

Today we are introducing Dyna-2, a world-action model pre-trained on one million hours of human video. At this scale, for the first time, we discovered several new scaling laws:

• world-action models exhibit scaling law on human data across four orders of magnitude, from 1000 to 1,000,000 hours,
• this human data scaling law implied a scaling law on never seen robot data,
• both data and objective matter; world modeling and scaling on video data are essential for cross-embodiment scaling transfer to emerge

🧵

[Chet Bhateja @ChetBhateja]

@Theonash_ @adcock_brett No

[Chet Bhateja @ChetBhateja]

@adcock_brett System 2?

[Chet Bhateja @ChetBhateja]

@HenrikMolgard @UCBerkeley I'm getting GPT-4o vibes.

[Joey Hejna @JoeyHejna]

As imitation learning policies continue to scale, deciding how to weigh different robot datasets will become even more difficult. To address this problem we introduce ReMix, a method for automatically curating large RT-X scale imitation learning datasets. 🧵(1/5)

[Chet Bhateja @ChetBhateja]

@ronawang wait till you hear about reverse passive income

[Chet Bhateja @ChetBhateja]

@MrBeast @LudwigAhgren and @Atrioc

[Chet Bhateja @ChetBhateja]

@MrBeast Some estimate in developing countries it takes a few thousand dollars to save a life so this could turn into a trolley problem...

[Chet Bhateja @ChetBhateja]

@gcodingcomps I'm not sure if this is better or worse lol.

[Chet Bhateja @ChetBhateja]

@gcodingcomps 1511? You gotta be kidding me.