The new engine moves that tooling lineage into robot physics simulation, where speed and generative scene creation can determine how many training iterations a research team can afford.
First-order effects
Robot-learning researchers and developers can use an open-source simulation system rather than relying solely on physical trials; Genesis claims simulated training can run up to 430,000 times faster than real-world training.
The immediate advantage is a much larger experiment budget for teams able to translate simulated results into robot behavior.
Second-order effects
Simulation platforms will compete less on raw access alone and more on physical fidelity, task diversity, and the ease of moving policies from simulation to hardware.
Faster environment generation complements automated training methods such as Eureka's reward-writing approach, potentially making the simulation-and-evaluation loop the key bottleneck rather than individual robot runs.
Third-order effects
If simulation-to-hardware transfer proves reliable, open-source physics engines could become shared infrastructure for robotics research, lowering the entry cost for training physical AI systems.
The durable constraint would shift from generating training experience to validating safety and performance on real machines, where simulation claims still require hardware evidence.
The trend: Robotics AI is moving toward scalable synthetic training stacks in which simulation, generated environments, and automated optimization reduce reliance on slow physical data collection.
This is roughly the equivalent of simulated protein folding, but for #robotics — This is also going to give video synthesis diffusion models (that work based on the statistical appearance of pixels) a run for their money, because this one is based on, you know, actual #physics.…
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 …
A cool demo showing soft robot simulation and control in Genesis (yes it's simulated), and then in real world :) made possible by amazing @johnsonwang0810 [video]
New physics sim trains robots 430,000 times faster than reality “Genesis” can compress training times from decades into hours using 3D worlds conjured from text. https://arstechnica.com/... [image]
20+ AI labs came together to build the most advanced physics engine ever created, and then they open-sourced it. Result: 10-80x faster than GPU stacks and 430,000x real-time. Consequence: Genesis is a generative physics engine: Text prompts turn into interactive 4D worlds, [video…
Genesis is a simulation platform for robot learning, supporting various materials, robots, and tasks, with generative and differentiable capabilities, and integrated with PyTorch for seamless neural network operations [image]
This is roughly the equivalent of simulated protein folding, but for #robotics Very impressive, very useful, and #opensource to boot. #AI #NeuralNetworks #ArtificialIntelligence https://arstechnica.com/...
Genesis is the first-ever platform providing comprehensive support for soft muscles and soft robot and their interaction with rigid robots. Genesis also ships with a URDF-like soft-robot configuration system. 4/n [video]
Genesis is an AI-enabled physics simulation platform designed for media, robotics, embodied AI, and physical AI applications. Genesis is a large scale research effort that spans over 20 labs, and is open-sourced. key features: ✦ universal physics engine ✦ photo-realistic [video]
have been working with the 𝐆𝐞𝐧𝐞𝐬𝐢𝐬 motion/animation system, and it is incredibly user-friendly! 💃🕺 It offers a suite of features for animation and motion control/generation, including character animation creation, simulation, and high-quality, visually striking rendering [video]
Open Source Genesis is going to completely change the personal robotics revolution and the many garage builders that will build upon incredible platforms like Tesla Optimus. My early experiment show this is a game changer! Link: https://github.com/... [video]
To put it in simple terms: Today with Open Source Genesis, you can train robots and of course, physics of games on the screen 430,000 faster than in real-time. It is the best AI thus far of this type. It will give OpenAI—headaches, this is FREE. [video]
Open Source Genesis is absolutely changing AI. The physics engine is 10-80x faster than existing GPU-accelerated stacks. It delivers a simulation speed ~430,000 faster than in real-time, and takes only 26 seconds to train a robotic locomotion. [video]
Once people start connecting the dots and realise the potential of the the team and tek behind #R3D @DebotScience there will be a huge repricing. Genesis is the latest AI Robotics project which was just announced yesterday on X via https://x.com/... This links back to a [image]
New physics sim trains robots 430,000 times faster than reality | “Genesis” can compress training times from decades into hours using 3D worlds conjured from text.