David Pantera
@davidpantera_ • 6,162 subscribers
Product @GoogleDeepMind (world models) | Scout @a16z | @Stanford 🌲 | Posting about AI | Views are mine!
Videos
Two of us filmed this on a couple phones, and we used Atlas to turn it into a short film. This is just a clip from the World Labs tech blog Atlas is able to generate the parts of the scene that none of the cameras actually filmed. It can freeze time, allowing you to reframe shots to positions never filmed, then resume. View events from impossible angles because Atlas knows what was there.
David Pantera951,950 Aufrufe • vor 1 Monat
With Atlas, you can literally make films. And it also basically solves first-frame-last-frame transitions. Reconstruct a 3D scene from images, then position the camera exactly where you want it, and let Atlas fill in what the camera would see along that path. You’re in the director's chair, staging the scene. Here we made a 1 min video with a couple images of totally different scenes, and we set where the camera keyframes should move through each part of the scene. The scene-to-scene transitions are great. Atlas makes it feel like a smooth shot with no cuts.
David Pantera27,621 Aufrufe • vor 1 Monat
Atlas is huge for real-to-sim. Scanning spaces like this usually requires tons of equipment, but Atlas can reconstruct the environment from just a couple pictures. We then simulate robots navigating through the generated world. Atlas generates what the robot would see from the perspective of its cameras. Even depth. It generates the environment AND the training data from it...
David Pantera13,704 Aufrufe • vor 1 Monat
We're not talking about world models enough. We're obsessed LLMs, but LLMs understand text, not reality. They are map-readers, not explorers. A robot arm doesn't have to break 1,000 actual cups to learn how to hold one. It simulates the grip, detects the slip, and adjusts its force in its internal world model first. It learns in milliseconds, safely. We don't just ask an AI to hypothesize based on what its been trained on; we use a world model to simulate how a theoretical drug molecule binds to a virus, accelerating discovery from years to minutes. This video is some clips of my favorite worlds I've been making in World Labs , who just opened up their private beta to the public this week. In the private beta, I spent hours spinning up worlds either from prompts or via uploading images (like this one of Roble gym @ Stanford). Some of the worlds here are public creations from other users, since you can post your worlds to a feed for others to explore in (it's a pretty fun social product, walking through worlds from other ppl's imaginations).
David Pantera29,280 Aufrufe • vor 11 Monaten
Keine weiteren Inhalte verfügbar