正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Pretty excited about the new SuperDex robotics simulator by Meta. 1. It's contact-first simulator in C++ designed for smooth, non-convex rigid–soft physics (vs mujoco designed for convex shapes) 2. Rather than standalone simulator, it includes a full suite of tools (RL/gym environments ("Superdex Lab"), GUI authoring tool ("Superdex Studio"),... show more

pfung

4,790 subscribers

15,286 次观看 • 21 天前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

5 条评论

vatnik 的头像

vatnik21 天前

@Meta Interesting, how well does it handle high deformations of material?

db 的头像

db21 天前

@Meta This looks amazing! Been working on a simpler version of sim2real and this adds a lot more excitement to it!

Nathaniel Nifong 的头像

Nathaniel Nifong21 天前

@Meta Strings!

satyam 的头像

satyam21 天前

@Meta Does it run in the browser over webrtc?

Chakradhari shiva 的头像

Chakradhari shiva21 天前

@Meta Interesting! @grok how does this differ from mujoco?

相关视频

Meta just open-sourced its dexterity stack! 🪬 Most physics simulators were built for things that move through space. Walking robots, drones, cars. Contact is the part they approximate worst, and obviously dexterous manipulation is nothing but contact. Project SuperDex, from Meta Reality Labs Research, is built the other way around, a contact-first physics engine with the whole platform stacked on top of it. The cool part is that it's on GitHub. The engine runs one solver across rigid bodies, soft bodies, rods and tendons, shells and cloth, in the same model. → Non-convex collision with accurate contact force distributions, so a multi-finger grasp gets simulated rather than approximated → Tactile sensors and soft contact as first-class primitives → Numerical stability without the tight time-step limits explicit solvers force on you → Constraint-aware inverse kinematics running on the same optimization core as the forward dynamics Then the data layer. Put on a Quest 3, teleoperate the simulated hand with haptic feedback, and generate demonstration datasets without touching real hardware. They show a shape-sorting policy trained entirely in simulation and deployed zero-shot on a real robotic hand. Robot hands are getting good. Data for them isn't that fast. Meta is betting the cheapest way to collect contact-rich demonstrations is a headset people already own, pointed at a simulator instead of a game. 🔗 Here's the project page: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Meta just open-sourced its dexterity stack! 🪬 Most physics simulators were built for things that move through space. Walking robots, drones, cars. Contact is the part they approximate worst, and obviously dexterous manipulation is nothing but contact. Project SuperDex, from Meta Reality Labs Research, is built the other way around, a contact-first physics engine with the whole platform stacked on top of it. The cool part is that it's on GitHub. The engine runs one solver across rigid bodies, soft bodies, rods and tendons, shells and cloth, in the same model. → Non-convex collision with accurate contact force distributions, so a multi-finger grasp gets simulated rather than approximated → Tactile sensors and soft contact as first-class primitives → Numerical stability without the tight time-step limits explicit solvers force on you → Constraint-aware inverse kinematics running on the same optimization core as the forward dynamics Then the data layer. Put on a Quest 3, teleoperate the simulated hand with haptic feedback, and generate demonstration datasets without touching real hardware. They show a shape-sorting policy trained entirely in simulation and deployed zero-shot on a real robotic hand. Robot hands are getting good. Data for them isn't that fast. Meta is betting the cheapest way to collect contact-rich demonstrations is a headset people already own, pointed at a simulator instead of a game. 🔗 Here's the project page: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Lukas Ziegler

60,286 次观看 • 21 天前

🧑‍🚀 Day 5 of the Cursor #vibejam Proudly sponsored by Cursor + bolt.new + GLIF Today a little bit about Cursor 3, it's a completely rebuilt interface designed around agents, you can run multiple AI agents in parallel, hand off between local and cloud, and go from commit to merged PR. It's powered by Composer 2, their own frontier coding model that tops benchmarks at $0.50/M input tokens — and the Fast variant runs at 200 tokens/sec, more than 2x faster than other frontier models. Perfect for vibe coding your game jam entry. And edwin from Cursor is here to help you if you get stuck! Lots of activity again today again on here, the best games (or well previews) I saw today: 🍣 Sushi Belt by steeno 🌊 Surfing Simulator by Erik (this one is a basic demo but very promising as it has realistic water physics) 🦆 Grand Quack Auto by kyzo (or as I like to call it Bali Simulator) 🏛️ Roman Combat Sim by terry trusner (continued from yesterday because it's had a lot of progress I think) YOU HAVE 24 DAYS LEFT! Reply in this thread with updates on your current games to share your progress, and add tag #vibejam so I see and can include you in the daily tweet There's $35,000 in prizes for you to win, see threads below for more info. The Gold prize is $20,000, bronze is $10,000 and silver is $5,000! Wanna to participate? You can still start today and submit your game any time before May 1!

🧑‍🚀 Day 5 of the Cursor #vibejam Proudly sponsored by Cursor + bolt.new + GLIF Today a little bit about Cursor 3, it's a completely rebuilt interface designed around agents, you can run multiple AI agents in parallel, hand off between local and cloud, and go from commit to merged PR. It's powered by Composer 2, their own frontier coding model that tops benchmarks at $0.50/M input tokens — and the Fast variant runs at 200 tokens/sec, more than 2x faster than other frontier models. Perfect for vibe coding your game jam entry. And edwin from Cursor is here to help you if you get stuck! Lots of activity again today again on here, the best games (or well previews) I saw today: 🍣 Sushi Belt by steeno 🌊 Surfing Simulator by Erik (this one is a basic demo but very promising as it has realistic water physics) 🦆 Grand Quack Auto by kyzo (or as I like to call it Bali Simulator) 🏛️ Roman Combat Sim by terry trusner (continued from yesterday because it's had a lot of progress I think) YOU HAVE 24 DAYS LEFT! Reply in this thread with updates on your current games to share your progress, and add tag #vibejam so I see and can include you in the daily tweet There's $35,000 in prizes for you to win, see threads below for more info. The Gold prize is $20,000, bronze is $10,000 and silver is $5,000! Wanna to participate? You can still start today and submit your game any time before May 1!

@levelsio

79,700 次观看 • 5 个月前

I DON’T THINK PEOPLE REALIZE HOW MASSIVE THIS IS YET Hugging Face and Pollen Robotics just released Microduck. It is a new legged robot for $399. It passed $1 million in sales in under SEVEN hours 🤯 The hardware packs 15 actuators, a camera, and LiDAR into a 25 cm frame. It weighs under 800 grams and runs for about an hour on a single charge. Out of the box, it can > walk > roller-skate > pick up objects > recover from falls .. and so much more! You can drive the 7 shipped policies from the first boot using a gamepad. But the real hook is the training stack. The software is entirely open: → MuJoCo simulation and PPO training → 50 Hz onboard policy loop → Browser simulator included alongside hardware → Full control stack available on GitHub The mechanical and electronic design files are still proprietary. But don’t let the cute robotic duck fool you. Under the closed chassis is a FULLY OPEN-SOURCE, hackable brain you can experiment with and build on. Repo in 🧵↓

I DON’T THINK PEOPLE REALIZE HOW MASSIVE THIS IS YET Hugging Face and Pollen Robotics just released Microduck. It is a new legged robot for $399. It passed $1 million in sales in under SEVEN hours 🤯 The hardware packs 15 actuators, a camera, and LiDAR into a 25 cm frame. It weighs under 800 grams and runs for about an hour on a single charge. Out of the box, it can > walk > roller-skate > pick up objects > recover from falls .. and so much more! You can drive the 7 shipped policies from the first boot using a gamepad. But the real hook is the training stack. The software is entirely open: → MuJoCo simulation and PPO training → 50 Hz onboard policy loop → Browser simulator included alongside hardware → Full control stack available on GitHub The mechanical and electronic design files are still proprietary. But don’t let the cute robotic duck fool you. Under the closed chassis is a FULLY OPEN-SOURCE, hackable brain you can experiment with and build on. Repo in 🧵↓

Charly Wargnier ♨️

651,063 次观看 • 18 天前

Everybody is talking about recursive self-improvement (RSI) and meta learning. Here is my old 2020 talk about this [1]. It has aged well. Example: humans still define the starts & ends of trials of many modern meta learners. My RSI systems since 1994 LEARN to (re)define them [2]! [1] Meta Learning Machines in a Single Lifelong Trial (talk for workshops at ICML 2020 and NeurIPS 2021, based on earlier talks since 1994). Abstract: the most widely used machine learning algorithms were designed by humans and thus are hindered by our cognitive biases and limitations. Can we also construct meta learning algorithms that can learn better learning algorithms so that our self-improving AIs have no limits other than those inherited from computability and physics? This question has been a main driver of my research since I wrote a thesis on it in 1987 [2]. Here I summarize our work on meta reinforcement learning with self-modifying policies in a single lifelong trial (since 1994), and mathematically optimal meta-learning through the self-referential Gödel Machine (since 2003). Many additional publications on meta-learning since 1987 can be found in the RSI overview [2]. [2] J. Schmidhuber (AI Blog, 2020-2025). 1/3 century anniversary of first publication on recursive self-improvement (RSI) and meta learning machines that learn to learn (1987). For its cover I drew a robot that bootstraps itself. 1992-: gradient descent-based neural meta learning. 1994-: meta reinforcement learning with self-modifying policies. 1997: meta RL plus artificial curiosity and intrinsic motivation. 2002-: asymptotically optimal meta learning for curriculum learning. 2003-: mathematically optimal Gödel Machine. 2020-: new stuff!

Everybody is talking about recursive self-improvement (RSI) and meta learning. Here is my old 2020 talk about this [1]. It has aged well. Example: humans still define the starts & ends of trials of many modern meta learners. My RSI systems since 1994 LEARN to (re)define them [2]! [1] Meta Learning Machines in a Single Lifelong Trial (talk for workshops at ICML 2020 and NeurIPS 2021, based on earlier talks since 1994). Abstract: the most widely used machine learning algorithms were designed by humans and thus are hindered by our cognitive biases and limitations. Can we also construct meta learning algorithms that can learn better learning algorithms so that our self-improving AIs have no limits other than those inherited from computability and physics? This question has been a main driver of my research since I wrote a thesis on it in 1987 [2]. Here I summarize our work on meta reinforcement learning with self-modifying policies in a single lifelong trial (since 1994), and mathematically optimal meta-learning through the self-referential Gödel Machine (since 2003). Many additional publications on meta-learning since 1987 can be found in the RSI overview [2]. [2] J. Schmidhuber (AI Blog, 2020-2025). 1/3 century anniversary of first publication on recursive self-improvement (RSI) and meta learning machines that learn to learn (1987). For its cover I drew a robot that bootstraps itself. 1992-: gradient descent-based neural meta learning. 1994-: meta reinforcement learning with self-modifying policies. 1997: meta RL plus artificial curiosity and intrinsic motivation. 2002-: asymptotically optimal meta learning for curriculum learning. 2003-: mathematically optimal Gödel Machine. 2020-: new stuff!

Jürgen Schmidhuber

244,411 次观看 • 6 个月前

AI researcher breaks down why your $5,000 Mac Studio runs local AI slower than benchmarks promise: "Performance isn't about the hardware - it's about three software layers stacked on top of it." in 15 minutes he tests MLX, Ollama, llama.cpp, and the new vllm-mlx head to head no GPU needed, just the Mac you probably already own - why Ollama, the most popular tool, is actually one of the slower options on Mac - the real bottleneck nobody benchmarks: 'prefill' time before the first token even appears - the exact quantization setting (4-bit) that's the best speed-to-quality tradeoff - why unified memory makes Macs uniquely good for big models compared to Nvidia GPUs complete guide to choosing the right hardware for local AI is below 👇

AI researcher breaks down why your $5,000 Mac Studio runs local AI slower than benchmarks promise: "Performance isn't about the hardware - it's about three software layers stacked on top of it." in 15 minutes he tests MLX, Ollama, llama.cpp, and the new vllm-mlx head to head no GPU needed, just the Mac you probably already own - why Ollama, the most popular tool, is actually one of the slower options on Mac - the real bottleneck nobody benchmarks: 'prefill' time before the first token even appears - the exact quantization setting (4-bit) that's the best speed-to-quality tradeoff - why unified memory makes Macs uniquely good for big models compared to Nvidia GPUs complete guide to choosing the right hardware for local AI is below 👇

rewind

207,513 次观看 • 2 个月前

Just built a bot that first runs hyper-realistic MiroFish swarm simulations on every upcoming Bitcoin and crypto event. And then agent instantly trades the real live markets on Polymarket, already printing $12,000+ per day in testing. Couldn't hold back after diving into MiroFish. Took the new god-tier agent behavior simulator from that Chinese college quant who coded it in 10 days, exploded GitHub to 23k+ stars and bagged $4.1M from Shanda overnight.. Paired it with OpenClaw (24/7 autonomous execution) + Claude Opus 4.6. And in one day built my first version of private Polymarket bot. Now it: -> spawns thousands of agents with real memory and personalities -> runs full GraphRAG swarm simulations modeling exactly how news, ETF flows, macro data, whale activity and sentiment will move Bitcoin price -> simulates thousands of possible futures specifically for Polymarket Bitcoin contracts -> detects where the crowd probability is mispriced on every crypto market and extracts the real edge -> auto-trades the edges instantly through OpenClaw the moment the opportunity appears Testing the bot + MiroFish based simulator live right now. First runs already printing hard. Meanwhile there's a real trader crushing with a similar stack imo, $321k all-time profit and 12k/day, 100% won on Bitcoin markets. Wallet: My own Polymarket profile + full trade logs drop later once I scale it hard. New meta just dropped, don't miss out! Check the guide and all info below.

slash1s

114,880 次观看 • 6 个月前

Can GPT-4 teach a robot hand to do pen spinning tricks better than you do? I'm excited to announce Eureka, an open-ended agent that designs reward functions for robot dexterity at super-human level. It’s like Voyager in the space of a physics simulator API! Eureka bridges the gap between high-level reasoning (coding) and low-level motor control. It is a “hybrid-gradient architecture”: a black box, inference-only LLM instructs a white box, learnable neural network. The outer loop runs GPT-4 to refine the reward function (gradient-free), while the inner loop runs reinforcement learning to train a robot controller (gradient-based). We are able to scale up Eureka thanks to IsaacGym, a GPU-accelerated physics simulator that speeds up reality by 1000x. On a benchmark suite of 29 tasks across 10 robots, Eureka rewards outperform expert human-written ones on 83% of the tasks by 52% improvement margin on average. We are surprised that Eureka is able to learn pen spinning tricks, which are very difficult even for CGI artists to animate frame by frame! Eureka also enables a new form of in-context RLHF, which is able to incorporate a human operator’s feedback in natural language to steer and align the reward functions. It can serve as a powerful co-pilot for robot engineers to design sophisticated motor behaviors. As usual, we open-source everything! Welcome you all to check out our video gallery and try the codebase today: Paper: Code: Deep dive with me: 🧵

Can GPT-4 teach a robot hand to do pen spinning tricks better than you do? I'm excited to announce Eureka, an open-ended agent that designs reward functions for robot dexterity at super-human level. It’s like Voyager in the space of a physics simulator API! Eureka bridges the gap between high-level reasoning (coding) and low-level motor control. It is a “hybrid-gradient architecture”: a black box, inference-only LLM instructs a white box, learnable neural network. The outer loop runs GPT-4 to refine the reward function (gradient-free), while the inner loop runs reinforcement learning to train a robot controller (gradient-based). We are able to scale up Eureka thanks to IsaacGym, a GPU-accelerated physics simulator that speeds up reality by 1000x. On a benchmark suite of 29 tasks across 10 robots, Eureka rewards outperform expert human-written ones on 83% of the tasks by 52% improvement margin on average. We are surprised that Eureka is able to learn pen spinning tricks, which are very difficult even for CGI artists to animate frame by frame! Eureka also enables a new form of in-context RLHF, which is able to incorporate a human operator’s feedback in natural language to steer and align the reward functions. It can serve as a powerful co-pilot for robot engineers to design sophisticated motor behaviors. As usual, we open-source everything! Welcome you all to check out our video gallery and try the codebase today: Paper: Code: Deep dive with me: 🧵

Jim Fan

2,675,676 次观看 • 2 年前

Japan Just Built a HouseBot You Control Without Speaking and It Changes Everything! Donut Robotics has officially unveiled its first bipedal humanoid, Cinnamon 1, and instead of focusing on louder voices or bigger motors, the company went in the opposite direction. Silence. Cinnamon 1 introduces what Donut Robotics calls Silent Gesture Control, a system that allows the humanoid to be guided using simple hand and finger movements rather than spoken commands. This approach feels especially well suited for real world environments where traditional voice control falls apart. Busy factory floors. Construction sites filled with constant noise. Even quiet indoor settings where voice commands feel awkward or intrusive. It also opens the door for far more accessible human robot interaction, particularly for users with impairments. While the current Cinnamon 1 hardware is built on an OEM platform, the intelligence driving it is where Donut Robotics is placing its long term bet. The team is actively developing custom Vision Language Action AI that allows the robot to interpret what it sees, understand intent, and respond with physical action. The goal is not just smarter robots, but robots that feel more natural. Even more ambitious is the company’s plan for full domestic production. Donut Robotics has stated its intention to localize both manufacturing and AI development in Japan, reinforcing the country’s reputation for precision engineering and thoughtful robotics design. If timelines hold, Cinnamon 1 units are expected to begin deployment in factories and construction environments by the end of 2026. That puts this humanoid squarely in the category of near term reality rather than distant concept. The takeaway is simple but important. As humanoid robots move out of labs and into daily work environments, the winners may not be the loudest or flashiest machines. They may be the ones that understand us without a word being spoken.

Japan Just Built a HouseBot You Control Without Speaking and It Changes Everything! Donut Robotics has officially unveiled its first bipedal humanoid, Cinnamon 1, and instead of focusing on louder voices or bigger motors, the company went in the opposite direction. Silence. Cinnamon 1 introduces what Donut Robotics calls Silent Gesture Control, a system that allows the humanoid to be guided using simple hand and finger movements rather than spoken commands. This approach feels especially well suited for real world environments where traditional voice control falls apart. Busy factory floors. Construction sites filled with constant noise. Even quiet indoor settings where voice commands feel awkward or intrusive. It also opens the door for far more accessible human robot interaction, particularly for users with impairments. While the current Cinnamon 1 hardware is built on an OEM platform, the intelligence driving it is where Donut Robotics is placing its long term bet. The team is actively developing custom Vision Language Action AI that allows the robot to interpret what it sees, understand intent, and respond with physical action. The goal is not just smarter robots, but robots that feel more natural. Even more ambitious is the company’s plan for full domestic production. Donut Robotics has stated its intention to localize both manufacturing and AI development in Japan, reinforcing the country’s reputation for precision engineering and thoughtful robotics design. If timelines hold, Cinnamon 1 units are expected to begin deployment in factories and construction environments by the end of 2026. That puts this humanoid squarely in the category of near term reality rather than distant concept. The takeaway is simple but important. As humanoid robots move out of labs and into daily work environments, the winners may not be the loudest or flashiest machines. They may be the ones that understand us without a word being spoken.

The AI Robot Guy on X

257,928 次观看 • 7 个月前

Let's reverse engineer Disney's adorable, lifelike robot! I couldn't find a whitepaper, but this is how I think it's trained: 1. The emotional behaviors are curated by Disney animation artists, keyframe by keyframe. But it cannot be "rendered" directly on the robot because it doesn't take into account the complex real-world physics. 2. Reinforcement learning (RL) is a great tool for training low-level robot controllers. RL needs a reward function to optimize, and it's typically a task reward (e.g. walk in a straight line as fast as possible). The problem is that RL doesn't know what counts as "natural behavior", and often produces weird-looking body postures that somehow still maximize the reward. This is a human alignment problem just like ChatGPT. 3. Enters Adversarial Motion Prior (AMP): a technique that learns the human preference by training a classifier on what we consider "emotional & cute". In GAN literature, this is called a discriminator. Disney artists are good at creating such a dataset. You can then add AMP as an auxiliary reward in simulation to nudge the robot towards desired behaviors. AMP was developed by Peng et al. 2021 and Escontrela et al. 2022. 4. Add lots of data augmentation to make the controller robust to physical disturbances. In RL, it's called "domain randomization". This is a very powerful technique that bridges the gap between simulator and reality. Previously, OpenAI used domain randomization to train a 5-finger robot hand to manipulate a Rubik's Cube: IEEE news article gave hints about the pipeline: Finally, praying for world peace 🙏. I hope robotics like this will bring more joy to the world.

Let's reverse engineer Disney's adorable, lifelike robot! I couldn't find a whitepaper, but this is how I think it's trained: 1. The emotional behaviors are curated by Disney animation artists, keyframe by keyframe. But it cannot be "rendered" directly on the robot because it doesn't take into account the complex real-world physics. 2. Reinforcement learning (RL) is a great tool for training low-level robot controllers. RL needs a reward function to optimize, and it's typically a task reward (e.g. walk in a straight line as fast as possible). The problem is that RL doesn't know what counts as "natural behavior", and often produces weird-looking body postures that somehow still maximize the reward. This is a human alignment problem just like ChatGPT. 3. Enters Adversarial Motion Prior (AMP): a technique that learns the human preference by training a classifier on what we consider "emotional & cute". In GAN literature, this is called a discriminator. Disney artists are good at creating such a dataset. You can then add AMP as an auxiliary reward in simulation to nudge the robot towards desired behaviors. AMP was developed by Peng et al. 2021 and Escontrela et al. 2022. 4. Add lots of data augmentation to make the controller robust to physical disturbances. In RL, it's called "domain randomization". This is a very powerful technique that bridges the gap between simulator and reality. Previously, OpenAI used domain randomization to train a 5-finger robot hand to manipulate a Rubik's Cube: IEEE news article gave hints about the pipeline: Finally, praying for world peace 🙏. I hope robotics like this will bring more joy to the world.

Jim Fan

314,694 次观看 • 2 年前

Special thanks to Google DeepMind for inviting me to try out Genie 3. I'm excited to share my thoughts on this early research prototype and also some of my live recordings below: I spent the whole day playing with the system and when it works, it is truly mind blowing🤯. It is the first neural game engine / world model I have tried that generalizes so well and has long term world consistency. Here’s a couple of examples from my live recording and some thoughts on what it means for the future of gaming, robotics, digital experiences and ASI. Where it shines: - Truly general-purpose and quick startup time. Works exceptionally well for gaming environments but also generalizes to other industrial and real-world scenarios. - It learns physics. Although there are systematic failures even for rigid body physics, it was clear to me that it can learn game engine and non-rigid physics without an underlying engine (and in limit learn from game engines via training data). - It works exceptionally well for stylized environments with characters walking around. This will have implications for concept artists, level designers and game devs. - It is way more fun than video models, indicating that there are high retention consumer experiences waiting to be built with this in the future - Photorealistic walk throughs and drone shots work exceptionally well - Global illumination and lighting works surprisingly well - Visual memory is quite powerful and the same objects approximately remain coherent under occlusion and longer time horizons Open Problems: - Physics is still hard and there are obvious failure cases when I tried the classical intuitive physics experiments from psychology (tower of blocks). - Social and multi-agent interactions are tricky to handle. 1vs1 combat games do not work - Long instruction following and simple combinatorial game logic fails (e.g. collect some points / keys etc, go to the door, unlock and so on) - Action space is limited - It is far from being a real game engines and has a long way to go but this is a clear glimpse into the future. The Future: - It is impressive enough for me to have strong conviction that this is going to disrupt the gaming industry. It is super early days and there are a lot of failures but the writing is on the wall. Lots of challenging scientific, engineering and scaling problems to be solved but it is going to happen in the next 5 years. - This is the final piece before we get full AGI and now I think we are well on our way to truly solve it once something like this is scaled up. In many ways it is more ASI than AGI but this is a matter of definitions. The fidelity and generalizability will reach human-level and quickly surpass humans - People are going to combine this with 3D AI and LLMs to build AAA games.

Special thanks to Google DeepMind for inviting me to try out Genie 3. I'm excited to share my thoughts on this early research prototype and also some of my live recordings below: I spent the whole day playing with the system and when it works, it is truly mind blowing🤯. It is the first neural game engine / world model I have tried that generalizes so well and has long term world consistency. Here’s a couple of examples from my live recording and some thoughts on what it means for the future of gaming, robotics, digital experiences and ASI. Where it shines: - Truly general-purpose and quick startup time. Works exceptionally well for gaming environments but also generalizes to other industrial and real-world scenarios. - It learns physics. Although there are systematic failures even for rigid body physics, it was clear to me that it can learn game engine and non-rigid physics without an underlying engine (and in limit learn from game engines via training data). - It works exceptionally well for stylized environments with characters walking around. This will have implications for concept artists, level designers and game devs. - It is way more fun than video models, indicating that there are high retention consumer experiences waiting to be built with this in the future - Photorealistic walk throughs and drone shots work exceptionally well - Global illumination and lighting works surprisingly well - Visual memory is quite powerful and the same objects approximately remain coherent under occlusion and longer time horizons Open Problems: - Physics is still hard and there are obvious failure cases when I tried the classical intuitive physics experiments from psychology (tower of blocks). - Social and multi-agent interactions are tricky to handle. 1vs1 combat games do not work - Long instruction following and simple combinatorial game logic fails (e.g. collect some points / keys etc, go to the door, unlock and so on) - Action space is limited - It is far from being a real game engines and has a long way to go but this is a clear glimpse into the future. The Future: - It is impressive enough for me to have strong conviction that this is going to disrupt the gaming industry. It is super early days and there are a lot of failures but the writing is on the wall. Lots of challenging scientific, engineering and scaling problems to be solved but it is going to happen in the next 5 years. - This is the final piece before we get full AGI and now I think we are well on our way to truly solve it once something like this is scaled up. In many ways it is more ASI than AGI but this is a matter of definitions. The fidelity and generalizability will reach human-level and quickly surpass humans - People are going to combine this with 3D AI and LLMs to build AAA games.

Tejas Kulkarni

88,083 次观看 • 1 年前

Demis Hassabis just told you why civilization never became a science. Hassabis: “AI itself will maybe unlock new sciences… the one I’m particularly excited about is AI for simulations.” Physics became physics because you could run the experiment twice. Drop a ball. Measure the fall. Drop it again. Same answer. Now you own the law. Economics never got that privilege. You raise interest rates on 300 million people and watch what breaks. Hassabis: “If you raise interest rates by half a percent, you have to do it in the real world and then see what happens. You can have theories, but you can’t run it thousands of times.” Every war. Every policy. Every financial system ever designed. One run. No control group. No second attempt. The population was the experiment and the cost. All of human history is a series of unrepeatable experiments performed on people who never consented to the trial. We buried the failures and called the survivors wise. Hassabis: “Why aren’t they just sciences like physics today? Because the problem is they’re emergent systems… it’s very hard to do repeated controlled experiments.” The line between hard science and soft science was never about intelligence. It was about whether you could afford to be wrong more than once. Physics could. Civilization could not. So we built governments on instinct. Economies on ideology. Foreign policy on pattern recognition one generation deep. And convinced ourselves that was rigor. Hassabis wants to end that era. Hassabis: “If you could simulate things really accurately, then maybe there’s sort of new sciences to be done where you can rigorously sample from a very accurate simulator.” Simulate a nation. Crash its economy. Isolate one variable. Run it again. A thousand iterations. A thousand variations. Before a single real person absorbs the cost. That is not a better forecasting tool. That is the end of governance by intuition. Hassabis: “That will allow us to make much better decisions in these, today, what are very uncertain domains.” Every field we called soft was only soft because the hardware to make it hard did not exist yet. He is not improving prediction. He is making civilization itself repeatable. And the moment it becomes repeatable, every lesson we thought we learned from history reveals itself for what it always was. A conclusion drawn from a sample size of one. That is not knowledge. That is mythology with better record-keeping.

Demis Hassabis just told you why civilization never became a science. Hassabis: “AI itself will maybe unlock new sciences… the one I’m particularly excited about is AI for simulations.” Physics became physics because you could run the experiment twice. Drop a ball. Measure the fall. Drop it again. Same answer. Now you own the law. Economics never got that privilege. You raise interest rates on 300 million people and watch what breaks. Hassabis: “If you raise interest rates by half a percent, you have to do it in the real world and then see what happens. You can have theories, but you can’t run it thousands of times.” Every war. Every policy. Every financial system ever designed. One run. No control group. No second attempt. The population was the experiment and the cost. All of human history is a series of unrepeatable experiments performed on people who never consented to the trial. We buried the failures and called the survivors wise. Hassabis: “Why aren’t they just sciences like physics today? Because the problem is they’re emergent systems… it’s very hard to do repeated controlled experiments.” The line between hard science and soft science was never about intelligence. It was about whether you could afford to be wrong more than once. Physics could. Civilization could not. So we built governments on instinct. Economies on ideology. Foreign policy on pattern recognition one generation deep. And convinced ourselves that was rigor. Hassabis wants to end that era. Hassabis: “If you could simulate things really accurately, then maybe there’s sort of new sciences to be done where you can rigorously sample from a very accurate simulator.” Simulate a nation. Crash its economy. Isolate one variable. Run it again. A thousand iterations. A thousand variations. Before a single real person absorbs the cost. That is not a better forecasting tool. That is the end of governance by intuition. Hassabis: “That will allow us to make much better decisions in these, today, what are very uncertain domains.” Every field we called soft was only soft because the hardware to make it hard did not exist yet. He is not improving prediction. He is making civilization itself repeatable. And the moment it becomes repeatable, every lesson we thought we learned from history reveals itself for what it always was. A conclusion drawn from a sample size of one. That is not knowledge. That is mythology with better record-keeping.

Dustin

23,348 次观看 • 2 个月前

When Christopher Nolan explains the plot of Momento, he draws a hairpin. It’s an idea I keep coming back to: Stories as shapes. In Momento, half of the movie moves forward in time while the other half moves backward. Then, toward the end, the two sides meet near the tip of the hairpin. Point 1: Nolan says to “approach structure mathematically with diagrams.” In storytelling, the idea of structure is everywhere: • 3 Act • StoryGrid • Hero’s Journey But Nolan claims those restrict his creativity. He prefers to think in ‘shapes’ rather than in hard rules. “What I do is draw many diagrams,” Nolan says, “Particularly if there’s a type of structural complexity.” Point 2: I’m reminded of Kurt Vonnegut, who taught about the “shapes of stories” at Harvard and Case Western. He broke down stories into ~8 unique shapes. Each runs on 2 axes: • X – Time • Y – Good vs ill fortune For example, Nolan’s plot map of Inception looks like Vonnegut’s “Man In Hole” shape. But Momento is non-linear so it doesn’t follow any of the 8. And yet Nolan still describes it as a shape. *** “Breaking rules isn’t interesting. It’s making up new ones that keeps things exciting.” – Christopher Nolan Follow Nathan Baugh for more ideas like this!

When Christopher Nolan explains the plot of Momento, he draws a hairpin. It’s an idea I keep coming back to: Stories as shapes. In Momento, half of the movie moves forward in time while the other half moves backward. Then, toward the end, the two sides meet near the tip of the hairpin. Point 1: Nolan says to “approach structure mathematically with diagrams.” In storytelling, the idea of structure is everywhere: • 3 Act • StoryGrid • Hero’s Journey But Nolan claims those restrict his creativity. He prefers to think in ‘shapes’ rather than in hard rules. “What I do is draw many diagrams,” Nolan says, “Particularly if there’s a type of structural complexity.” Point 2: I’m reminded of Kurt Vonnegut, who taught about the “shapes of stories” at Harvard and Case Western. He broke down stories into ~8 unique shapes. Each runs on 2 axes: • X – Time • Y – Good vs ill fortune For example, Nolan’s plot map of Inception looks like Vonnegut’s “Man In Hole” shape. But Momento is non-linear so it doesn’t follow any of the 8. And yet Nolan still describes it as a shape. *** “Breaking rules isn’t interesting. It’s making up new ones that keeps things exciting.” – Christopher Nolan Follow Nathan Baugh for more ideas like this!

Nathan Baugh

256,319 次观看 • 3 年前

Everything you love about generative models — now powered by real physics! Announcing the Genesis project — after a 24-month large-scale research collaboration involving over 20 research labs — a generative physics engine able to generate 4D dynamical worlds powered by a physics simulation platform designed for general-purpose robotics and physical AI applications. Genesis's physics engine is developed in pure Python, while being 10-80x faster than existing GPU-accelerated stacks like Isaac Gym and MJX. It delivers a simulation speed ~430,000 faster than in real-time, and takes only 26 seconds to train a robotic locomotion policy transferrable to the real world on a single RTX4090 (see tutorial: The Genesis physics engine and simulation platform is fully open source at We'll gradually roll out access to our generative framework in the near future. Genesis implements a unified simulation framework all from scratch, integrating a wide spectrum of state-of-the-art physics solvers, allowing simulation of the whole physical world in a virtual realm with the highest realism. We aim to build a universal data engine that leverages an upper-level generative framework to autonomously create physical worlds, together with various modes of data, including environments, camera motions, robotic task proposals, reward functions, robot policies, character motions, fully interactive 3D scenes, open-world articulated assets, and more, aiming towards fully automated data generation for robotics, physical AI and other applications. Open Source Code: Project webpage: Documentation: 1/n

Everything you love about generative models — now powered by real physics! Announcing the Genesis project — after a 24-month large-scale research collaboration involving over 20 research labs — a generative physics engine able to generate 4D dynamical worlds powered by a physics simulation platform designed for general-purpose robotics and physical AI applications. Genesis's physics engine is developed in pure Python, while being 10-80x faster than existing GPU-accelerated stacks like Isaac Gym and MJX. It delivers a simulation speed ~430,000 faster than in real-time, and takes only 26 seconds to train a robotic locomotion policy transferrable to the real world on a single RTX4090 (see tutorial: The Genesis physics engine and simulation platform is fully open source at We'll gradually roll out access to our generative framework in the near future. Genesis implements a unified simulation framework all from scratch, integrating a wide spectrum of state-of-the-art physics solvers, allowing simulation of the whole physical world in a virtual realm with the highest realism. We aim to build a universal data engine that leverages an upper-level generative framework to autonomously create physical worlds, together with various modes of data, including environments, camera motions, robotic task proposals, reward functions, robot policies, character motions, fully interactive 3D scenes, open-world articulated assets, and more, aiming towards fully automated data generation for robotics, physical AI and other applications. Open Source Code: Project webpage: Documentation: 1/n

Zhou Xian

3,821,556 次观看 • 1 年前

New "superman" Four months later, Shanghai-based Matrix Robotics unveiled the Matrix-3--yet another anthropomorphic humanoid robot situated on the left side of the "Uncanny Valley." Unlike Boston Dynamics' Atlas, which is designed for industrial settings, the Matrix-3 is a general-purpose humanoid robot intended initially for service-oriented environments such as supermarkets, hotels, and office buildings. Standing 1.7 meters tall and weighing 65 kilograms, the adult-sized robot features a full-body covering of biomimetic fabric integrated with tactile sensors, as well as humanoid biomimetic muscles. It employs linear joints rather than the mainstream rotary joints. Matrix claims it is capable of pulling or pushing a load of 200 kg. Optimus makes extensive use of linear joints (linear actuators). It is equipped with dexterous hands offering 27 DOF and boasts a battery life of 4 hours. The robot is powered by a model named WAVE(like an E2E World Action Model) Matrix states that it has already established a manufacturing facility in Shanghai's Zhangjiang with an annual production capacity of 10,000 units. The company plans to deliver 1,000 units this year, with the first batch scheduled for delivery by the end of June. Pricing starts at $99,000. Notably, Matrix founder Zhang Haixing previously served as a leader at Tesla China's Design and Research Center, where he was deeply involved in the early-stage R&D and design of the Optimus robot. This company seems to be delivering on the future envisioned for Optimus.

New "superman" Four months later, Shanghai-based Matrix Robotics unveiled the Matrix-3--yet another anthropomorphic humanoid robot situated on the left side of the "Uncanny Valley." Unlike Boston Dynamics' Atlas, which is designed for industrial settings, the Matrix-3 is a general-purpose humanoid robot intended initially for service-oriented environments such as supermarkets, hotels, and office buildings. Standing 1.7 meters tall and weighing 65 kilograms, the adult-sized robot features a full-body covering of biomimetic fabric integrated with tactile sensors, as well as humanoid biomimetic muscles. It employs linear joints rather than the mainstream rotary joints. Matrix claims it is capable of pulling or pushing a load of 200 kg. Optimus makes extensive use of linear joints (linear actuators). It is equipped with dexterous hands offering 27 DOF and boasts a battery life of 4 hours. The robot is powered by a model named WAVE(like an E2E World Action Model) Matrix states that it has already established a manufacturing facility in Shanghai's Zhangjiang with an annual production capacity of 10,000 units. The company plans to deliver 1,000 units this year, with the first batch scheduled for delivery by the end of June. Pricing starts at $99,000. Notably, Matrix founder Zhang Haixing previously served as a leader at Tesla China's Design and Research Center, where he was deeply involved in the early-stage R&D and design of the Optimus robot. This company seems to be delivering on the future envisioned for Optimus.

CyberRobo

40,727 次观看 • 4 个月前

Real quick thoughts on Bellum - which entered Beta yesterday. 1) There is nothing else like it. It's the only game designed specifically to force teamwork in a hardcore PVP tactical shooter. It does it in new ways: SA tools locked to leadership roles, novel respawn mechanics requiring you to maneuver near teammates and punishes lone wolves, and more. 2) It's a game about gunfights, and it does them so damn well. Fire and maneuver works. Fire superiority matters. Cover, concealment, obscuration, even IMTs are rewarded while bonzai charges are nearly always fatal. 3) I think what speaks to me most is how well it models the cognitive overload of a maneuver leader. A Squad leader in this game is monitoring/working up to 2 nets + local at once - Command, Squad, and VOIP. This is modeled in other games, of course. But Bellum adds some spice - your hand comes off your weapon to activate the PTT, so you can't fire particularly well or do medical tasks while working a net. By default, you have platoon assault net in one ear and squad assault in the other. It feels like I'm wearing peltors for an FMP or going on target. But despite all of the stimulus, leaders have to stay locked in to win, painting the picture for their squads, reporting higher, marking the map, and getting shit done despite the chaos. It's in day 2 of beta, so I'll hold my criticisms for now. But this game is on track to do some things we've not seen before, and to do them exceptionally well. Very excited to see where this one goes.

Real quick thoughts on Bellum - which entered Beta yesterday. 1) There is nothing else like it. It's the only game designed specifically to force teamwork in a hardcore PVP tactical shooter. It does it in new ways: SA tools locked to leadership roles, novel respawn mechanics requiring you to maneuver near teammates and punishes lone wolves, and more. 2) It's a game about gunfights, and it does them so damn well. Fire and maneuver works. Fire superiority matters. Cover, concealment, obscuration, even IMTs are rewarded while bonzai charges are nearly always fatal. 3) I think what speaks to me most is how well it models the cognitive overload of a maneuver leader. A Squad leader in this game is monitoring/working up to 2 nets + local at once - Command, Squad, and VOIP. This is modeled in other games, of course. But Bellum adds some spice - your hand comes off your weapon to activate the PTT, so you can't fire particularly well or do medical tasks while working a net. By default, you have platoon assault net in one ear and squad assault in the other. It feels like I'm wearing peltors for an FMP or going on target. But despite all of the stimulus, leaders have to stay locked in to win, painting the picture for their squads, reporting higher, marking the map, and getting shit done despite the chaos. It's in day 2 of beta, so I'll hold my criticisms for now. But this game is on track to do some things we've not seen before, and to do them exceptionally well. Very excited to see where this one goes.

ControlledPairs

61,458 次观看 • 5 个月前

This might be the best "AI Engineer" I've tried so far. I'm an old school developer who started 30 years ago. I feel very uncomfortable letting AI take control of my code, but for the sake of science, I spent two hours building an application that took me weeks to build a couple of years ago. I used Pythagora, a brand new tool backed by Y Combinator. They just released to the public. Keep in mind that I use AI every day to write code, but Pythagora is something different: it's a tool that leads, and uses you—the human—as the copilot. I go into more details in the video, but here is the TLDR; 1. Holy molly! We've made a ton of progress on this front! This is way better than Devin when I tested it a few months back. 2. Love the approach of generating a plan with sub-tasks before writing any code. 3. The tools never tries to do too much: it tackles every small task one at a time, and gives you instructions so you test everything. 4. It does exactly what you'd do when it gets stuck: writes a bunch of logs and uses those to correct itself. Pretty neat! 5. It's fast. It runs locally. It's an extension to Visual Studio Code. I'm impressed, but I don't think this tool is for me. I'm not the type of developer who's ready to relinquish control. I felt I had no connection with the code because I didn't write it. It was not my code. I know many people who don't care about this. I know many people who will get tremendous value out of Pythagora. I hope they keep pushing the limits, providing feedback, and helping this get to a point where old folks like me feel more comfortable using it. Don't take my word for it. The best thing you can do is to give it a try and see how you feel using it. Thanks to the team who built this, for all of the explanations and support, and especially, for sitting and listening to my dumb questions for 2 hours while I tested this.

Santiago

212,056 次观看 • 1 年前

I modelled the head, face, and battle mask for Optimus Prime for #Transformers ROTB. I also designed his new "centurion" open battle mask look along with it's animation. The director wanted the face to resemble the face from the 2007 Transformers film as ROTB was a soft-prequel to it but also wanted it to be tweaked to fit the new design and for us to give it some new unique features. This is where we came up with the new mask design which doesn't get entirely hidden when open. Instead, it gets stored on the side of his cheeks, like a Roman centurions helmet. The head design is derived from his model from the 2018 Bumblebee film to keep some consistency with that film too as ROTB was also a soft-sequel to Bumblebee. There are new pieces, adjusted proportions, and reworks to detailing in ROTBs version. The face proportions are modified from the 2007 face to be a lot wider and beefier, matching his new head shape and look. I redesigned and built the eyes from scratch using our kitbash pieces like I did for all the Maximals and Mirage. The studio decided against going for the holographic eyes look from the 2018 Bumblebee film but rather wanted a more simple look in between the 2007 and the 2018 eyes. I added a lot of small mechanical details to his battle mask animation which we unfortunately never get to see in the film. For example, I designed a fully functional mechanism which pushes his chin piece back when his mask closes. This was very cool when it was visible, our Animation Director and VFX Supervisors were super excited about it, but unfortunately no shot showed it off in the final film. It was a lot of fun to help model and design Optimus Prime for this film. I'm happy I got to contribute to such an iconic character. CREDITS: Video: Paramount, MPC, WetaFX Ruairidh MacNeill modelled the body for Optimus

I modelled the head, face, and battle mask for Optimus Prime for #Transformers ROTB. I also designed his new "centurion" open battle mask look along with it's animation. The director wanted the face to resemble the face from the 2007 Transformers film as ROTB was a soft-prequel to it but also wanted it to be tweaked to fit the new design and for us to give it some new unique features. This is where we came up with the new mask design which doesn't get entirely hidden when open. Instead, it gets stored on the side of his cheeks, like a Roman centurions helmet. The head design is derived from his model from the 2018 Bumblebee film to keep some consistency with that film too as ROTB was also a soft-sequel to Bumblebee. There are new pieces, adjusted proportions, and reworks to detailing in ROTBs version. The face proportions are modified from the 2007 face to be a lot wider and beefier, matching his new head shape and look. I redesigned and built the eyes from scratch using our kitbash pieces like I did for all the Maximals and Mirage. The studio decided against going for the holographic eyes look from the 2018 Bumblebee film but rather wanted a more simple look in between the 2007 and the 2018 eyes. I added a lot of small mechanical details to his battle mask animation which we unfortunately never get to see in the film. For example, I designed a fully functional mechanism which pushes his chin piece back when his mask closes. This was very cool when it was visible, our Animation Director and VFX Supervisors were super excited about it, but unfortunately no shot showed it off in the final film. It was a lot of fun to help model and design Optimus Prime for this film. I'm happy I got to contribute to such an iconic character. CREDITS: Video: Paramount, MPC, WetaFX Ruairidh MacNeill modelled the body for Optimus

Rassoul Edji

388,573 次观看 • 1 年前

Alex Proyas on the Commercial failure of 'Dark City' (1998) & its similarities with 'The Matrix' (1999): "Interviewer: Are you surprised by how big a cult film it has become? Proyas: I was surprised it didn't achieve commercial success initially, but I'm happy it has had this long-lasting life after the fact. I was pretty disappointed the studio hadn't really managed to sell the movie in the way that they could have. I don't think they really knew what to make of it quite frankly. The studio put out a trailer that was quite clever. It was completely non-narrative, just music and imagery. The guys who made the trailer specifically wrote a piece of music for the trailer, and the whole thing felt handcrafted and definitely had a personality. I think people were genuinely impressed by it but it didn't explain to them what the movie was or give them an idea of the kind of experience they were going to get. I think it made audiences think it was some kind of a horror movie. As a result people kind of overlooked it in the theaters. I know 99% of the people who loved it and always ask me about it are people who saw it on video years later. How do you feel about the similarities the later 'The Matrix' (1999) has to 'Dark City' (1998)? If I answer that question honestly it sounds like sour grapes because obviously 'The Matrix' did better commercially than 'Dark City' did. 'Dark City' came out a year before 'The Matrix', one of my producers on 'Dark City' worked on 'The Matrix', and the Wachowskis saw 'Dark City' a good nine months before 'Dark City' came out in the theaters, so they had a good run at picking the more important cherries off the cake. I feel like it's a question I can't answer in any more detail than that, other than to say that maybe the Wachowskis should be asked the question rather than me." (Alex Proyas' interview with Paul Rowlands, 2017)

Alex Proyas on the Commercial failure of 'Dark City' (1998) & its similarities with 'The Matrix' (1999): "Interviewer: Are you surprised by how big a cult film it has become? Proyas: I was surprised it didn't achieve commercial success initially, but I'm happy it has had this long-lasting life after the fact. I was pretty disappointed the studio hadn't really managed to sell the movie in the way that they could have. I don't think they really knew what to make of it quite frankly. The studio put out a trailer that was quite clever. It was completely non-narrative, just music and imagery. The guys who made the trailer specifically wrote a piece of music for the trailer, and the whole thing felt handcrafted and definitely had a personality. I think people were genuinely impressed by it but it didn't explain to them what the movie was or give them an idea of the kind of experience they were going to get. I think it made audiences think it was some kind of a horror movie. As a result people kind of overlooked it in the theaters. I know 99% of the people who loved it and always ask me about it are people who saw it on video years later. How do you feel about the similarities the later 'The Matrix' (1999) has to 'Dark City' (1998)? If I answer that question honestly it sounds like sour grapes because obviously 'The Matrix' did better commercially than 'Dark City' did. 'Dark City' came out a year before 'The Matrix', one of my producers on 'Dark City' worked on 'The Matrix', and the Wachowskis saw 'Dark City' a good nine months before 'Dark City' came out in the theaters, so they had a good run at picking the more important cherries off the cake. I feel like it's a question I can't answer in any more detail than that, other than to say that maybe the Wachowskis should be asked the question rather than me." (Alex Proyas' interview with Paul Rowlands, 2017)

DepressedBergman

66,180 次观看 • 8 个月前