←
Bidhan Roy2,526 posts
Bidhan Roy avatar

Bidhan Roy

@bidhan

ceo @bageldotcom. building general world model for physical AI.

◷ Joined January 2021▧ 143 Photos and videos
Bidhan Roy avatar
Bidhan Roy@bidhan·Oct 14

This fully autonomous demo shows a Bagel Labs foundation model controlling a robot to make coffee. We previously released WorldDiT, our lightweight architecture for robot foundation models. At the time of release, it was on the Pareto frontier for model size and performance. We have since been refining this architecture to handle longer horizon, more useful tasks fully autonomously. Making coffee means getting a whole sequence of actions right. Each step depends on the result of the last, so small mistakes can build up as the task continues. Getting this to work with a small model is especially challenging. This demo is an example of what the model can do today. If you are deploying robots or have existing hardware you want to make autonomous, we would like to partner with you. bagel.com/partner

1:32
avi avatar
⟳ Bidhan Roy retweeted
avi@avizurlo·Nov 10

anthropic pivoting from chart crime to demo crime

Post attachment

Fable 5.2 shows substantially improved skills in controlling physical robots, even on tasks it has never explicitly been trained to do. On RoboDojo, a benchmark with a simulated two-arm robot, it outperformed other models and most AI systems built to control robots.

0:23
bagel.com avatar
⟳ Bidhan Roy retweeted
bagel.com@bageldotcom·Nov 5

A relevant talk if you want to understand which industries are adopting Physical AI today!

I will be talking about physical AI at @WebSummit lisbon center stage next week. Anyone there next week interested in the same topics, please HMU!

Post attachment
Bidhan Roy avatar
Bidhan Roy@bidhan·Nov 5

I will be talking about physical AI at @WebSummit lisbon center stage next week. Anyone there next week interested in the same topics, please HMU!

Post attachment
Sam avatar
⟳ Bidhan Roy retweeted
Sam@futurenomics·Oct 31

Some predictions: >Increasingly likely that Anthropic will postpone IPO or numbers will come substantially below target (200b on 2T) >Data markets will get hit harder than compute markets by the cool down >Both will likely IPO by end of 2028. Too large to stay private

Post attachment

The ChatGPT maker’s latest funding round values it at $1.25 trillion, less than its $1.4 trillion target. on.wsj.com/4A55X3a

Charly Wargnier ♨️ avatar
⟳ Bidhan Roy retweeted
Charly Wargnier ♨️@DataChaz·Oct 14

SO MUCH FIREPOWER COMING OUT OF THIS NEW COMPANY 👀 @bageldotcom is building lightweight robotics foundation models, and they're already demonstrating fully autonomous, multi-step tasks. Impressive stuff!

1:32

This fully autonomous demo shows a Bagel Labs foundation model controlling a robot to make coffee. We previously released WorldDiT, our lightweight architecture for robot foundation models. At the time of release, it was on the Pareto frontier for model size and performance. We have since been refining this architecture to handle longer horizon, more useful tasks fully autonomously. Making coffee means getting a whole sequence of actions right. Each step depends on the result of the last, so small mistakes can build up as the task continues. Getting this to work with a small model is especially challenging. This demo is an example of what the model can do today. If you are deploying robots or have existing hardware you want to make autonomous, we would like to partner with you. bagel.com/partner

1:32
Bidhan Roy avatar
Bidhan Roy@bidhan·Oct 19

a SOTA world action model demo

This fully autonomous demo shows a Bagel Labs foundation model controlling a robot to make coffee. We previously released WorldDiT, our lightweight architecture for robot foundation models. At the time of release, it was on the Pareto frontier for model size and performance. We have since been refining this architecture to handle longer horizon, more useful tasks fully autonomously. Making coffee means getting a whole sequence of actions right. Each step depends on the result of the last, so small mistakes can build up as the task continues. Getting this to work with a small model is especially challenging. This demo is an example of what the model can do today. If you are deploying robots or have existing hardware you want to make autonomous, we would like to partner with you. bagel.com/partner

1:32
Bidhan Roy avatar
Bidhan Roy@bidhan·Oct 16

this. and that’s why meta as a frontier lab should formidable to all the other labs.

Someone asked me this week: "Why are data companies making so much money right now?" The answer is actually quite interesting, and speaks to how frontier labs don't have access to what is most important for the next era of intelligence: product surface.

a16z avatar
⟳ Bidhan Roy retweeted
a16z@a16z·Oct 12

optimism wins

Post attachment
PARSA avatar
⟳ Bidhan Roy retweeted
PARSA@Parsats_eth·Oct 14

Coffee is a surprisingly good test for a robot. Lots of small steps and each one has to go right for the next one to work

This fully autonomous demo shows a Bagel Labs foundation model controlling a robot to make coffee. We previously released WorldDiT, our lightweight architecture for robot foundation models. At the time of release, it was on the Pareto frontier for model size and performance. We have since been refining this architecture to handle longer horizon, more useful tasks fully autonomously. Making coffee means getting a whole sequence of actions right. Each step depends on the result of the last, so small mistakes can build up as the task continues. Getting this to work with a small model is especially challenging. This demo is an example of what the model can do today. If you are deploying robots or have existing hardware you want to make autonomous, we would like to partner with you. bagel.com/partner

1:32
Bidhan Roy avatar
Bidhan Roy@bidhan·Oct 14

pov

Post attachment

This fully autonomous demo shows a Bagel Labs foundation model controlling a robot to make coffee. We previously released WorldDiT, our lightweight architecture for robot foundation models. At the time of release, it was on the Pareto frontier for model size and performance. We have since been refining this architecture to handle longer horizon, more useful tasks fully autonomously. Making coffee means getting a whole sequence of actions right. Each step depends on the result of the last, so small mistakes can build up as the task continues. Getting this to work with a small model is especially challenging. This demo is an example of what the model can do today. If you are deploying robots or have existing hardware you want to make autonomous, we would like to partner with you. bagel.com/partner

1:32
Emerson S avatar
⟳ Bidhan Roy retweeted
Emerson S@Em_Nomadic·Oct 14

A cup of coffee is a surprisingly good test of robot intelligence. So many little decisions, movements, and opportunities for things to go wrong.

This fully autonomous demo shows a Bagel Labs foundation model controlling a robot to make coffee. We previously released WorldDiT, our lightweight architecture for robot foundation models. At the time of release, it was on the Pareto frontier for model size and performance. We have since been refining this architecture to handle longer horizon, more useful tasks fully autonomously. Making coffee means getting a whole sequence of actions right. Each step depends on the result of the last, so small mistakes can build up as the task continues. Getting this to work with a small model is especially challenging. This demo is an example of what the model can do today. If you are deploying robots or have existing hardware you want to make autonomous, we would like to partner with you. bagel.com/partner

1:32
Santiago avatar
⟳ Bidhan Roy retweeted
Santiago@svpino·Oct 14

My wife would absolutely love a robot that makes her coffee 3 times every day. I hope she doesn't see this.

This fully autonomous demo shows a Bagel Labs foundation model controlling a robot to make coffee. We previously released WorldDiT, our lightweight architecture for robot foundation models. At the time of release, it was on the Pareto frontier for model size and performance. We have since been refining this architecture to handle longer horizon, more useful tasks fully autonomously. Making coffee means getting a whole sequence of actions right. Each step depends on the result of the last, so small mistakes can build up as the task continues. Getting this to work with a small model is especially challenging. This demo is an example of what the model can do today. If you are deploying robots or have existing hardware you want to make autonomous, we would like to partner with you. bagel.com/partner

1:32