Odyssey

Odyssey is (always) hiring (in Zurich, London, and Palo Alto), but this one is especially close to my heart (and urgent): Scaling seems to work extremely well for world models just as it does for text models, yet world models are even more flops heavy and to control a robot or power interactive gaming experiences they have to run in real time. If you cherish the jax scaling book, live in perfetto, like the expressiveness of kernel-DSLs like Helion, and/or are curious to cover both nVidia and beyond (Trainium, AMD, etc.), we'll get along swimmingly. Link below to apply or share with friends who might be interested 🙂

It's time to build a new kind of superintelligence! One that learns more like humans do, gains knowledge through experience, and recursively improves by acting in and learning from the world. Our early results of this at Odyssey are so encouraging. Come shape it with us. https://lnkd.in/g62eXK4b

A new kind of superintelligence is being built, grown inside a world model. https://lnkd.in/gZuzqiic

We're hiring on all fronts!

We just closed a $310M Series B at Odyssey, and we're growing rapidly all across the company. I'm recruiting for several critical leadership roles, specifically Research TLMs, Head of Policy, and Head of Product. This is a very special, defining moment in world models, and I can't get across in words just how exciting it feels to be at Odyssey right now. It's incredible. Apply, meet with us, and I'd love to share more.

We’ve raised a $310M Series B to accelerate world models! We believe AI that can understand and simulate the world will be one of the most important technologies of our time. We're excited to partner with Natural Capital, Amazon, GV, AMD, IQT, and others to bring this to life. Natural Capital led this round, and we're thrilled to partner with such an experienced, high-conviction DeepTech investor! It's become clear that general world models will unlock entirely new capabilities across robotics, science, healthcare, education, gaming, and beyond. This capital and compute enables us to explore each of them, and to go beyond language models. In addition, we're partnering closely on research and compute with Amazon, who share our conviction that world models are within reach of a GPT-3 moment. We're exploring what becomes possible when AI can deeply understand and simulate the world. Join us!

World models can now create imagined experiences for AI—environments where agents continuously learn, adapt, and improve. We suspect multi-agent interaction may be a critical ingredient for recursive AI and general intelligence. https://lnkd.in/g7vj_-pX

Introducing Agora-1, a multi-agent world model. Multiple participants—human or AI—can now interact inside the same world simulation, all in real-time. Try our playable research preview today, with Agora-1 simulating a multiplayer GoldenEye deathmatch!

Meet our new friend, Starchild-1 ❤️ Starchild-1 is the first ever real-time multimodal world model. A world model understands and simulates the world. Starchild-1 has learned to generate not just the visuals of the world, but the sounds of it too!

Introducing PROWL! We’ve built RL agents that explore game environments, tasked with discovering failures in world models across physics, visuals, and actions. Those failures then become training data in an automated loop that advances world model performance. In Minecraft, PROWL unlocked major improvements in action following, visual quality, temporal coherence, environmental dynamics, and out-of-distribution robustness. It even learned to simulate entirely new behaviors discovered by RL agents. We see PROWL as another lever of scale for world models, which we're excited to explore further.