Karol Hausman's banner

Karol Hausman

@hausman_k • 43,088 subscribers

@Physical_int

Shorts

Very exited to announce our largest deep RL deployment to date: robots sorting trash end-to-end in real offices! (aka RLS) This project took a long time (started before SayCan/RT-1/other newer works) but the learnings from it have been really valuable.🧵

Very exited to announce our largest deep RL deployment to date: robots sorting trash end-to-end in real offices! (aka RLS) This project took a long time (started before SayCan/RT-1/other newer works) but the learnings from it have been really valuable.🧵

243,854 views

PaLM-E or GPT-4 can speak in many languages and understand images. What if they could speak robot actions? Introducing RT-2: our new model that uses a VLM (up to 55B params) backbone and fine-tunes it to directly output robot actions!

PaLM-E or GPT-4 can speak in many languages and understand images. What if they could speak robot actions? Introducing RT-2: our new model that uses a VLM (up to 55B params) backbone and fine-tunes it to directly output robot actions!

182,789 views

There are many cool subtasks that π-0.5 can perform (yes, all in unseen homes) Here are a few examples. Folding and hanging a towel:

There are many cool subtasks that π-0.5 can perform (yes, all in unseen homes) Here are a few examples. Folding and hanging a towel:

56,473 views

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

Sometimes things don't go as planned...

Sometimes things don't go as planned...

276,201 views • 1 year ago

We developed a general recipe that allows VLAs to improve from experience. RL is back. (yes, this is 13 hours of coffee making)

We developed a general recipe that allows VLAs to improve from experience. RL is back. (yes, this is 13 hours of coffee making)

111,355 views • 8 months ago

Sometimes it's not humans that mess with the robots, it's the real world. Two pieces of clothing on the table:

Sometimes it's not humans that mess with the robots, it's the real world. Two pieces of clothing on the table:

151,833 views • 1 year ago

PaLM-E or GPT-4 can speak in many languages and understand images. What if they could speak robot actions? Introducing RT-2: our new model that uses a VLM (up to 55B params) backbone and fine-tunes it to directly output robot actions!

PaLM-E or GPT-4 can speak in many languages and understand images. What if they could speak robot actions? Introducing RT-2: our new model that uses a VLM (up to 55B params) backbone and fine-tunes it to directly output robot actions!

182,789 views • 3 years ago

A few examples of π0.7 in action Turn the shirt right-side out and fold it:

A few examples of π0.7 in action Turn the shirt right-side out and fold it:

24,192 views • 3 months ago

Treat yourself to a 4-hour timelapse of a robot assembling boxes off-site:

Treat yourself to a 4-hour timelapse of a robot assembling boxes off-site:

27,485 views • 8 months ago

Seems like people find these 1x videos soothing, here is 50 mins of uninterrupted autonomous coffee making

Seems like people find these 1x videos soothing, here is 50 mins of uninterrupted autonomous coffee making

25,411 views • 8 months ago

Congrats on the Gemini 3 release Real AGI will need to be physical:

Congrats on the Gemini 3 release Real AGI will need to be physical:

23,211 views • 8 months ago

Many things can go wrong when running very long-horizon policies. One way to test that is by messing with the robot and seeing if it can recover. We've done quite a bit of that for π-0.5 testing:

Many things can go wrong when running very long-horizon policies. One way to test that is by messing with the robot and seeing if it can recover. We've done quite a bit of that for π-0.5 testing:

27,869 views • 1 year ago

No more content to load