Загрузка видео...

Не удалось загрузить видео

На главную

Can any human RL engineers come up with a reward function corresponding to "shadow boxing humanoid" or "b boy spinning spider"? It turns out that AI agents can do 😊

Kangwook_Lee's profile picture

Kangwook Lee

6,566 subscribers

41,623 просмотров • 4 месяцев назад •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

it turns out you can do everything AI can do with a green screen
0:18

it turns out you can do everything AI can do with a green screen

fahmeedx

48,479 просмотров • 5 месяцев назад

You can summarize any webpage with Firecrawl 🔥 Perfect for research, content curation, or building AI agents that need quick context! Try it out in our Playground, SDK, or your favorite no-code automation platform!
0:12

You can summarize any webpage with Firecrawl 🔥 Perfect for research, content curation, or building AI agents that need quick context! Try it out in our Playground, SDK, or your favorite no-code automation platform!

Firecrawl

13,731 просмотров • 7 месяцев назад

I found out the other day that any compression tool can be contorted to do language modeling. Turns out gzip can generate text that somewhat *resembles* Shakespeare. Short write up linked below
0:11

I found out the other day that any compression tool can be contorted to do language modeling. Turns out gzip can generate text that somewhat *resembles* Shakespeare. Short write up linked below

Nathan Barry

253,544 просмотров • 1 месяц назад

These are not CGI. Reinforcement learning is so back. When operating on strings, it gives us o3. When operating on physical motors, it gives us a perfect humanoid backflip and a robot creature that out-maneuvers almost every animal on earth. RL is one of the only learning algorithms that can master both the world of bits and the world of atoms. Give me a reward function, and I shall move the world. 2025, Year of RL.
0:12

These are not CGI. Reinforcement learning is so back. When operating on strings, it gives us o3. When operating on physical motors, it gives us a perfect humanoid backflip and a robot creature that out-maneuvers almost every animal on earth. RL is one of the only learning algorithms that can master both the world of bits and the world of atoms. Give me a reward function, and I shall move the world. 2025, Year of RL.

Jim Fan

356,800 просмотров • 1 год назад

"You don't have to speak Python, or C++ or Fortran. You just can speak human." AI is wiping out all the gates that once stood in the way to build and leverage technology. How much faster would the world progress with 100M software engineers vs 2M ?
0:16

"You don't have to speak Python, or C++ or Fortran. You just can speak human." AI is wiping out all the gates that once stood in the way to build and leverage technology. How much faster would the world progress with 100M software engineers vs 2M ?

Rohan Paul

96,836 просмотров • 8 месяцев назад

You can now generate public endpoints, UI snippets, and a MCP server with AI Search: making it easy to add search to your website or connect AI agents.
0:12

You can now generate public endpoints, UI snippets, and a MCP server with AI Search: making it easy to add search to your website or connect AI agents.

Cloudflare Developers

35,570 просмотров • 4 месяцев назад

You can now Coldlink any $ETH or $SOL wallet to a Roblox username. That means Roblox game developers can now easily recognise & reward you based on your holdings or on-chain history with zero risk (1/2)
0:15

You can now Coldlink any $ETH or $SOL wallet to a Roblox username. That means Roblox game developers can now easily recognise & reward you based on your holdings or on-chain history with zero risk (1/2)

Coldlink.xyz

15,439 просмотров • 1 год назад

Alpha boxing fraternity pledge - Seedance 2.0 I'm looking for a website community that was similar to the now defunct Sora.chatgpt to post my Seedance AI boixng videos so I can interact with other ai creators with similar interests. Any ideas?
0:15

Alpha boxing fraternity pledge - Seedance 2.0 I'm looking for a website community that was similar to the now defunct Sora.chatgpt to post my Seedance AI boixng videos so I can interact with other ai creators with similar interests. Any ideas?

Boxtp

11,383 просмотров • 2 месяцев назад

You can now build Cursor agents in Rust, Go, or any other language! We're open-sourcing the SDK Bridge so you can write a thin adapter in your language that keeps up when we add agent features. Try it at
0:19

You can now build Cursor agents in Rust, Go, or any other language! We're open-sourcing the SDK Bridge so you can write a thin adapter in your language that keeps up when we add agent features. Try it at

eric zakariasson

58,658 просмотров • 6 дней назад

Some fun knife play I added for Jane. You can spin the knife around and have it switch position or you can even throw it up while its spinning and she'll pick it up with her tail. Should be a lot of fun to play with in VR! #vrchatavatar #JaneDoe
0:14

Sensitive content

Some fun knife play I added for Jane. You can spin the knife around and have it switch position or you can even throw it up while its spinning and she'll pick it up with her tail. Should be a lot of fun to play with in VR! #vrchatavatar #JaneDoe

eN7ity 🐉

28,316 просмотров • 2 лет назад

🚀 We've just shipped our new AI-powered research agent Deploy it in any Attio workflow to automate nuanced tasks that typically require human judgment, like prospecting, lead qualification, and more. Whether you're PLG or sales-led, here's a sample of what you can do:
0:15

🚀 We've just shipped our new AI-powered research agent Deploy it in any Attio workflow to automate nuanced tasks that typically require human judgment, like prospecting, lead qualification, and more. Whether you're PLG or sales-led, here's a sample of what you can do:

Attio

11,249 просмотров • 2 лет назад

AI now can copy any ads just upload any existing ad and your product photo, it generate an product ad with realistic taking character holding your product and introducing it both video and audio are so real that no one can tell its AI here's how to do it:
0:14

AI now can copy any ads just upload any existing ad and your product photo, it generate an product ad with realistic taking character holding your product and introducing it both video and audio are so real that no one can tell its AI here's how to do it:

el.cine

81,592 просмотров • 8 месяцев назад

Well, a humanoid robot can move on two legs or wheels, depending on the environment it faces. AGIBOT (Zhiyuan) from Shanghai revealed that their new generation of humanoid robot "Nezha" can autonomously switch to two wheels for fast movement when walking on two legs. Imagine that in the airport hall, when he walks down the stairs and sees a fallen object not far away, it can transform into a wheeled robot and move quickly to pick it up. Or in elderly care, when it see an elderly person fall, it can quickly move and help the elderly person up, and quickly provide medical supplies to the caregiver.
0:10

Well, a humanoid robot can move on two legs or wheels, depending on the environment it faces. AGIBOT (Zhiyuan) from Shanghai revealed that their new generation of humanoid robot "Nezha" can autonomously switch to two wheels for fast movement when walking on two legs. Imagine that in the airport hall, when he walks down the stairs and sees a fallen object not far away, it can transform into a wheeled robot and move quickly to pick it up. Or in elderly care, when it see an elderly person fall, it can quickly move and help the elderly person up, and quickly provide medical supplies to the caregiver.

CyberRobo

40,245 просмотров • 1 год назад

250912 Kai x Esquire Korea Case File Q: Is there any newly coined term you've come up with or learned recently? 🐻 History wasn't made in a day, it's not like I can come up with something on the spot because you say "make a meme!" There's pains that have to come with creation.
0:13

250912 Kai x Esquire Korea Case File Q: Is there any newly coined term you've come up with or learned recently? 🐻 History wasn't made in a day, it's not like I can come up with something on the spot because you say "make a meme!" There's pains that have to come with creation.

INTLKAI

22,981 просмотров • 11 месяцев назад

omg... it's over... Hollywood has officially been left behind you can swap out any actor from any film in one click using Higgsfield’s Wan Replace, it even works with intense camera motion now, anyone can achieve Hollywood level motion control using AI here’s how to do it:
0:14

omg... it's over... Hollywood has officially been left behind you can swap out any actor from any film in one click using Higgsfield’s Wan Replace, it even works with intense camera motion now, anyone can achieve Hollywood level motion control using AI here’s how to do it:

el.cine

2,493,538 просмотров • 10 месяцев назад

Here's a look at Human Torch and Storm's Team up Ability "Omega Fire" Storm can consume one of Human Torch’s fire tornadoes into her ultimate to ignite her hurricane OR Human Torch can turn Storm’s hurricane into a fiery hurricane with his ultimate, which will then enlarge and intensify any of his fire tornadoes caught in its path.
0:11

Here's a look at Human Torch and Storm's Team up Ability "Omega Fire" Storm can consume one of Human Torch’s fire tornadoes into her ultimate to ignite her hurricane OR Human Torch can turn Storm’s hurricane into a fiery hurricane with his ultimate, which will then enlarge and intensify any of his fire tornadoes caught in its path.

Marvel Rivals Intel

982,761 просмотров • 1 год назад

New research from Databricks: LLMs Can Learn to Reason via Off-Policy RL Optimal Advantage-based Policy Optimization with Lagged Inference policy (OAPL) shows you don’t need strict on-policy training to improve reasoning. It matches or beats Group Relative Policy Optimization (GRPO), stays stable with large policy lag, and uses ~3× fewer training generations. For Databricks customers, it’s a simpler, practical, and equally powerful approach to RL that Databricks is pioneering internally — and bringing directly to Databricks customers, so enterprises can improve agents using the same methods we use for our in-house agents, without complex infrastructure changes.
0:13

New research from Databricks: LLMs Can Learn to Reason via Off-Policy RL Optimal Advantage-based Policy Optimization with Lagged Inference policy (OAPL) shows you don’t need strict on-policy training to improve reasoning. It matches or beats Group Relative Policy Optimization (GRPO), stays stable with large policy lag, and uses ~3× fewer training generations. For Databricks customers, it’s a simpler, practical, and equally powerful approach to RL that Databricks is pioneering internally — and bringing directly to Databricks customers, so enterprises can improve agents using the same methods we use for our in-house agents, without complex infrastructure changes.

Databricks AI Research

12,678 просмотров • 5 месяцев назад

China's Alibaba just announced Animate Anyone 2. This AI can take any single image of a character and animates it with realistic motion seamless to its environment. 10 wild examples: 1. Human Interaction
0:11

China's Alibaba just announced Animate Anyone 2. This AI can take any single image of a character and animates it with realistic motion seamless to its environment. 10 wild examples: 1. Human Interaction

Min Choi

1,017,517 просмотров • 1 год назад

I have never seen this before, can Bayonetta mains explain how I got slowed down here LOL My only guess is that it has something to do with the witch timed shadow ball and me catching up to it
0:16

I have never seen this before, can Bayonetta mains explain how I got slowed down here LOL My only guess is that it has something to do with the witch timed shadow ball and me catching up to it

Sillintor

20,690 просмотров • 2 лет назад

I just open-sourced my /learn skill. Learn anything with agents and HTML artifacts. I have been learning about all kinds of topics with it. Install the skill and interact with any agent to help you through any topic. Ask it to generate visual and interactive artifacts and help you go deeper or generate knowledge checks (e.g., quizzes). Upskilling myself on any topic is one of the most impactful ways I have been able to use AI agents. If you are a DAIR Academy pro member, you can use it with our AI Builder. Skill: Try now:
0:19

I just open-sourced my /learn skill. Learn anything with agents and HTML artifacts. I have been learning about all kinds of topics with it. Install the skill and interact with any agent to help you through any topic. Ask it to generate visual and interactive artifacts and help you go deeper or generate knowledge checks (e.g., quizzes). Upskilling myself on any topic is one of the most impactful ways I have been able to use AI agents. If you are a DAIR Academy pro member, you can use it with our AI Builder. Skill: Try now:

elvis

34,407 просмотров • 1 месяц назад