Lindon Gao(@Lindon_Gao):Robotics is entering its scaling law era. Dyna-2 isn’t a one-off demo. It addresses one of the biggest constraints in robot intelligence today: data. We show that as human experience scales from 1,000 to 1,000,000 hours, robot intelligence keeps improving, including on robots and tasks the model never saw during pre-training. More data. Better generalization. More capability. More use cases and deployments. And we’re already seeing capabilities emerge with scale that surprised us. Dyna-2 isn’t the only breakthrough we’ve been working on. There’s a lot more coming in the next few weeks.

Lindon Gao

@Lindon_Gao

Cofounder & CEO @DynaRobotics, Instacart VPE, Caper AI ($350m exit)

参加 November 2015

40 フォロー中 2.1K ファン

Lindon Gao@Lindon_Gao

2026.08.10 17:49

Robotics is entering its scaling law era. Dyna-2 isn’t a one-off demo. It addresses one of the biggest constraints in robot intelligence today: data. We show that as human experience scales from 1,000 to 1,000,000 hours, robot intelligence keeps improving, including on robots and tasks the model never saw during pre-training. More data. Better generalization. More capability. More use cases and deployments. And we’re already seeing capabilities emerge with scale that surprised us. Dyna-2 isn’t the only breakthrough we’ve been working on. There’s a lot more coming in the next few weeks.

112

コミュニティへ転送

人気のあるユーザー