Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Trying to teach my new model to run using reinforcement learning (Hyfydy + depRL)

Thomas Geijtenbeek

3,624 subscribers

975,960 views • 2 years ago •via X (Twitter)

Education Science & Technology Health & Wellness

Anya Rossi• Live Now

Private livecam show

9 Comments

Christopher Fryant2 years ago

Say what you will, this model's got style.

aizk ✡️2 years ago

Wait, is your software the one responsible for this amazing meme?

Thomas Geijtenbeek2 years ago

Haha, that was a long time ago. Didn’t know that meme though.

No apple no life 🇺🇦 🇮🇱2 years ago

You’re welcome 🤭

Frank Bruno2 years ago

Wait... did it really? 🤣😅😅🤣🤣

Brian Jordan2 years ago

That looks so much like the first couple weeks of walking attempts by toddlers or like me playing qwop

SirisysAlpha2 years ago

Great work! May I offer "Spiral Line: The Human bodies internal jump rope of bilateral stabilization". . . It may provide your model with the needed additional axes of control and stabilization. + @POINTBLANK_LLC

Rob Ritchie2 years ago

Bro taught it the thriller

James Clark2 years ago

I wonder if you could sell these "weird walks". My first thought was "that would look great for zombies in videogames". Maybe you could sell the movement as a game asset?

Related Videos

DeepMind training bipedal robots to play soccer using deep reinforcement learning.

DeepMind training bipedal robots to play soccer using deep reinforcement learning.

MachinePix

16,988,388 views • 3 years ago

notes on my reinforcement learning for rocket league side quest so far: i’m in awe that — we live in a time where i can train a reinforcement learning model that is learning to play rocket league and at the same time, test community made rocket league models. where, both processes can run in parallel & locally on my hardware. in the game window — i’m playing against a reinforcement learning ppo model that’s running locally. and in the vscode window — i’m training my own reinforcement learning model (just a dummy example script for now). also, the rocket league botting ecosystem has incredible tooling where you can test out different reinforcement learning ideas against other algorithms + it has great documentation for linux and windows. it’s honestly a beautiful hidden gem of a rabbit hole :)

notes on my reinforcement learning for rocket league side quest so far: i’m in awe that — we live in a time where i can train a reinforcement learning model that is learning to play rocket league and at the same time, test community made rocket league models. where, both processes can run in parallel & locally on my hardware. in the game window — i’m playing against a reinforcement learning ppo model that’s running locally. and in the vscode window — i’m training my own reinforcement learning model (just a dummy example script for now). also, the rocket league botting ecosystem has incredible tooling where you can test out different reinforcement learning ideas against other algorithms + it has great documentation for linux and windows. it’s honestly a beautiful hidden gem of a rabbit hole :)

naklecha

46,795 views • 1 year ago

.Thomas Geijtenbeek shared a new simulation demonstrating real-time interaction with a biomechanically accurate model created using #SCONE and #Hyfydy. Learn more: #simulation #machinelearning

.Thomas Geijtenbeek shared a new simulation demonstrating real-time interaction with a biomechanically accurate model created using #SCONE and #Hyfydy. Learn more: #simulation #machinelearning

80 LEVEL

11,658 views • 1 year ago

New Model + Learning how to animate

New Model + Learning how to animate

𝐁𝐢𝐥𝐬𝐮 | wishlist Trinity 64 plz

34,412 views • 11 months ago

What if we could teach an AI to master the strategic game of 2048 through pure reinforcement learning? I did exactly that with "Agent 2048" - fine-tuning Qwen 7B model using GRPO to develop spatial reasoning and merge strategies with zero prior gameplay SFTdata! Thanks to Hugging Face and Unsloth AI for their easy to use implementation kalomaze and will brown you might like this :)

What if we could teach an AI to master the strategic game of 2048 through pure reinforcement learning? I did exactly that with "Agent 2048" - fine-tuning Qwen 7B model using GRPO to develop spatial reasoning and merge strategies with zero prior gameplay SFTdata! Thanks to Hugging Face and Unsloth AI for their easy to use implementation kalomaze and will brown you might like this :)

Hrishbh Dalal

31,863 views • 1 year ago

trying to teach deadlock to my friend group:

trying to teach deadlock to my friend group:

Cole (they finally buffed freja)

188,239 views • 3 months ago

Using our brain simulator, we’ve trained a reinforcement learning agent to maximize bits per second. Here is the RL policy converting brain data to cursor control in simulation:

Using our brain simulator, we’ve trained a reinforcement learning agent to maximize bits per second. Here is the RL policy converting brain data to cursor control in simulation:

Neuralink

18,410 views • 1 year ago

today we are teaching a machine to exploit object detection models using swarm drones + reinforcement learning after only 5,000 steps on my Mac Air M3, the policy brought down detection from 90% to 32%, allowing 20+ drones to fly undetected

today we are teaching a machine to exploit object detection models using swarm drones + reinforcement learning after only 5,000 steps on my Mac Air M3, the policy brought down detection from 90% to 32%, allowing 20+ drones to fly undetected

chester

180,750 views • 11 months ago

Who are you trying to teach my guy?

Who are you trying to teach my guy?

Dark Side

201,688 views • 9 months ago

Frontier models that use reasoning to "think" during inference are generating 5X more AI tokens per year. ✨ "Inference is now a thinking process. And in order to teach AI how to think, reinforcement learning and very significant computation was introduced into post-training," said NVIDIA CEO Jensen Huang in a recent keynote at #CES2026. Reinforcement learning is increasing computation demands across all AI scaling laws: pre-training, post-training, and test-time scaling. Learn more about reinforcement learning and AI scaling ➡️

Frontier models that use reasoning to "think" during inference are generating 5X more AI tokens per year. ✨ "Inference is now a thinking process. And in order to teach AI how to think, reinforcement learning and very significant computation was introduced into post-training," said NVIDIA CEO Jensen Huang in a recent keynote at #CES2026. Reinforcement learning is increasing computation demands across all AI scaling laws: pre-training, post-training, and test-time scaling. Learn more about reinforcement learning and AI scaling ➡️

NVIDIA AI Infrastructure

13,329 views • 6 months ago

Trying to teach the huzz why I have the best daytrading model in the world

Trying to teach the huzz why I have the best daytrading model in the world

Ryan

32,926 views • 4 months ago

Robot policies must be both reliable and highly capable to be useful; the best way to achieve this level of performance is with reinforcement learning. However, for reinforcement learning you are usually stuck between two difficult options: reinforcement in the real world is often risky and expensive, while reinforcement learning in a traditional simulator takes a lot of engineering work and has a persistent sim-to-real gap. What if instead you could train your robot purely in a world model? RISE by Jiazhi Yang et al. uses a compositional world model to predict the future and evaluate progress. This allows for a self-improving pipeline, which learns a world model from real data and then learns how the robot should perform different tasks. This pipeline results in a data-driven way to improve policy performance from real data but without real-world reinforcement learning. Watch Episode #86 of RoboPapers, with Chris Paxton and Jiafei Duan, to learn more!

Robot policies must be both reliable and highly capable to be useful; the best way to achieve this level of performance is with reinforcement learning. However, for reinforcement learning you are usually stuck between two difficult options: reinforcement in the real world is often risky and expensive, while reinforcement learning in a traditional simulator takes a lot of engineering work and has a persistent sim-to-real gap. What if instead you could train your robot purely in a world model? RISE by Jiazhi Yang et al. uses a compositional world model to predict the future and evaluate progress. This allows for a self-improving pipeline, which learns a world model from real data and then learns how the robot should perform different tasks. This pipeline results in a data-driven way to improve policy performance from real data but without real-world reinforcement learning. Watch Episode #86 of RoboPapers, with Chris Paxton and Jiafei Duan, to learn more!

RoboPapers

38,334 views • 1 month ago

OpenAI's Sebastien Bubeck on the o1 paradigm: "No tactic was given to the model. Everything is emergent. Everything is learned through reinforcement learning. This is insane. Insanity"

OpenAI's Sebastien Bubeck on the o1 paradigm: "No tactic was given to the model. Everything is emergent. Everything is learned through reinforcement learning. This is insane. Insanity"

Paul Calcraft

267,679 views • 1 year ago

Sexy soft steps.Adam learned to walk like a human,trained by reinforcement learning.

Sexy soft steps.Adam learned to walk like a human,trained by reinforcement learning.

CyberRobo

19,732 views • 1 year ago

Training AI to Play Pokemon with Reinforcement Learning by Peter Whidden github: youtube:

Training AI to Play Pokemon with Reinforcement Learning by Peter Whidden github: youtube:

AK

837,837 views • 2 years ago

robot named CyberRunner, developed by researchers at ETH Zurich, which uses Al and model-based reinforcement learning to master the classic Labyrinth board game in just six hours

robot named CyberRunner, developed by researchers at ETH Zurich, which uses Al and model-based reinforcement learning to master the classic Labyrinth board game in just six hours

Science girl

120,161 views • 11 months ago

Excited to partner with Cassyni to generate citable DOIs for my videos!!! Example: if you want to cite my video on Reinforcement Learning DOI Youtube Check out the complete collection

Excited to partner with Cassyni to generate citable DOIs for my videos!!! Example: if you want to cite my video on Reinforcement Learning DOI Youtube Check out the complete collection

Steven Brunton

36,044 views • 3 years ago

Screenshot to HTML Keeps trying to match your screenshot to the HTML output using OpenAI vision model

Screenshot to HTML Keeps trying to match your screenshot to the HTML output using OpenAI vision model

Dushyant

2,165,992 views • 2 years ago

how to learn something from the first principles (with a live demo on reinforcement learning)

how to learn something from the first principles (with a live demo on reinforcement learning)

Ajeet ( opensox.ai )

232,836 views • 2 months ago

trying to make a cyber arm, hard surface is not my home but we are learning new things 😊😊

trying to make a cyber arm, hard surface is not my home but we are learning new things 😊😊

ZinPia

12,373 views • 6 months ago