The new AI model of Odyssey flow 3D interactive

OdysseyAn young company founded by self -driving pioneers Oliver Cameron and Jeff HookIt has developed the Amnesty International model that allows users to “interact” with the flowing video.

Available on the web in the “Early Employee”, the model creates and flows video tires every 40 milliliters. With the basic controls, viewers can explore the areas inside a video clip, similar to a 3D video game.

“Given the current situation of the world, a possible work, and a history of countries and procedures, the model tries to predict the following situation in the world,” Odessi explains in a blog post. “This is a new global model, which shows capabilities such as generating pixel units that feel realistic, maintaining spatial consistency, video learning procedures, and removing coherent video flows for 5 minutes or more.”

A number of startups and large technology companies are chasing after that World modelsIt includes DeepmindFei-FEI Lee’s Impact Intelligence Researcher World laboratoriesand MicrosoftAnd Decart. They believe that world models can be used one day to create interactive media, such as games and movies, and play realistic simulations such as robots training environments.

But designs have mixed feelings about technology. A recently WIRED’s investigation found that games studios like Activision Blizzard, which resulted in dozens of workers, used artificial intelligence to cut corners and combat attrition. And 2024 Ticket Encouraged by the Animation Syndicate, the union, which represents the animation in Hollywood and cartoonists, has estimated that more than 100,000 movies, TV and animation functions will be disrupted by artificial intelligence in the coming months.

For its part, Odyssey pledged to cooperate with creative professionals – not to replace them.

“The interactive video (…) opens the door to completely new forms of entertainment, where stories can be created and explored upon request, free of restrictions and traditional production costs,” the company writes in the blog post. “With the passage of time, we believe that everything is today’s video – entertainment, ads, education, training, travel and more – will develop into an interactive video, all supported by Odysse.”

The Odyssey’s illustration is a little bit on the edges, which the company recognizes in its publication. The environments generated by the model are blurry, dispersed, and unstable in the sense that their layouts do not always remain the same. Walking forward in one direction for a period of time or turning, and the surrounding may seem suddenly different.

But the promising company quickly improves the form, which can currently broadcast the video up to 30 frames of NVIDIA H100 GPU at a cost of 1-2 dollars per “user watch”.

“We look forward to the future, we are looking for the richest representations of the world that embodies the dynamics much more honest, while increasing temporal stability and the ongoing situation,” Odyssey wrote in office. “Parallel, we expand the work area from the movement to the global interaction, and we learn open procedures from the video on a large scale.”

Odyssey takes a different approach from many artificial intelligence laboratories in the global modeling space. Designed a 360 degrees, a back camera system To capture the landscape in the real world, which is believed to Odyssey can serve as a basis for high -quality models of only trained models on the data available to the public.

To date, Odyssey has raised $ 27 million of investors including EQT Ventures, GV and Air Street Capital. Located Catmull, one of the founders of Pixar and former head of the Walt Disney Mobile Studios, on the startup board.

Last December, Odysse He said He was working on programs that allow creators to download scenes that were created by their models to tools such as Unreal Engine, Blender and Adobe After so that they could be manually edited.