Google DeepMind dropped Genie 3, a new world model image / video AI that can create open worlds where you can move around like in a video game in real-time. Step towards the holodeck.

in Deutsch D-A-CH4 days ago (edited)

Google DeepMind hat eine neue Bild/Video-KI vorgestellt, die Open-Worlds mit einem Prompt generieren kann, in denen man sich wie in einem Videogame in Echtzeit bewegen kann.

Genie 3 kann in 720p 24 Bilder pro Sekunde rendern, was beeindruckend ist und es ist auch möglich mit der Umgebung zu interagieren und mit Prompts Veränderungen vorzunehmen.

Das Besondere ist auch, dass sich die KI einige Minuten lang die Umgebung und die durchgeführten Änderungen merken kann (World Memory), was zur guten Konsistenz der generierten Bilder beiträgt.

Manche spekulieren, dass mit dieser Technologie das Erstellen von Videogames revolutioniert werden könnte. Anstatt aufwendig 3D-Grafiken manuell zu erstellen, könnte in Zukunft eine KI die gesamte Welt generieren.

Was Genie 3 noch nicht kann, ist Menschen/NPCs zu generieren, aber das wird vermutlich auch irgendwann möglich sein.

Jedenfalls ist es ein gewaltiger Schritt Richtung Star-Trek-Holodeck. Mit Prompts wird man sich in Zukunft ganze Welten in Echtzeit nach den eigenen Vorlieben erstellen und darin leben können. Fast so wie auf dem Holodeck von Star Trek.

Was sagt ihr dazu?

Demos of Genie 3

image.png

~~~ embed:1952732150928724043 twitter metadata:R29vZ2xlRGVlcE1pbmR8fGh0dHBzOi8vdHdpdHRlci5jb20vR29vZ2xlRGVlcE1pbmQvc3RhdHVzLzE5NTI3MzIxNTA5Mjg3MjQwNDN8 ~~~

~~~ embed:1952732999193096392 twitter metadata:anBhcmtlcmhvbGRlcnx8aHR0cHM6Ly90d2l0dGVyLmNvbS9qcGFya2VyaG9sZGVyL3N0YXR1cy8xOTUyNzMyOTk5MTkzMDk2MzkyfA== ~~~

Genie 3: A new frontier for world models

https://deepmind.google/discover/blog/genie-3-a-new-frontier-for-world-models/

English

Google DeepMind has unveiled a new image/video AI that can generate open worlds with a prompt, allowing users to move around in real-time as if in a video game.

Genie 3 can render 24 frames per second in 720p, which is impressive, and it is also possible to interact with the environment and make changes using prompts.

What's also impressive is that the AI can remember the environment and the changes made for a few minutes (World Memory), which contributes to the exceptional consistency of the generated images.

Some speculate that this technology could revolutionize the creation of video games. Instead of tediously creating complex 3D graphics by hand, an AI could generate the entire world in the future.

What Genie 3 cannot yet do is generate humans/NPCs, but that will probably be possible at some point as well.

It is a huge step towards the Star Trek holodeck. In the future, prompts will allow you to create entire worlds in real-time according to your own preferences and live in them. Almost like the holodeck in Star Trek.

Sort:  

Scary innovation but super cool and somehow needed, nonetheless. I can see it being very useful in the architectural realm, too.

AI is improving fast, and it's amazing to see how amazing the things you can create are.

Graphic designers in the unemployment queue in 3, 2, 1

could always find a job mining lithium in Africa for our green e-car batteries - but the competition there is fierce. https://www.miningreview.com/energy/top-10-lithium-producers-in-africa/
Lithium Mine in Africa

Crazy to see the evolution of ai these years

Aber ein Drink wie im Holodeck, den du trinken kannst wirds leider nicht geben.
Hatte die Tage Picard Staffel 3 gebingt.

Die Technik schreitet ab jetzt glaube in riesen Schritten voran, wo wir wohl i. 10 Jahren sind?

!BBH

We’re getting closer to living inside our imaginations. Once it can do NPCs, it’s basically game over for traditional game design.

Würde das gerne einmal testen, sobald es verfügbar wird.

This technology is evolving at breakneck speed.

In the future you will be able to play a different video game every time and without needing to buy it as Ai will make it

!PIZZA

Ich habe soeben auf Gemini ein KI Bild kreirt:

Gemini_Generated_Image_kaxx9ukaxx9ukaxx.jpg
das war mein prompt:
Surrealist oil painting. Extreme close-up. A grotesque, anthropomorphic aardvark ringmaster with a baboon's head and wearing sunglasses stares at a self-portrait. The sunglasses reflect them. The portrait is a Hieronymus Bosch-inspired surrealist artwork. Octane render. Unreal Engine 5. Glowing colors. Detailed brushstrokes. Maximalist. Triadic colors. Professional photography. 8k resolution. A masterpiece. Artstation.

PIZZA!

$PIZZA slices delivered:
@davideownzall(4/15) tipped @vikisecrets

Come get MOONed!

That is a huge milestone for Google that will help shape the future

!BBH

This is absolutely mind-blowing. Genie 3 feels like a major leap toward the sci-fi future we've only dreamed about, like the Star Trek holodeck coming to life. Real-time, prompt-based generation of interactive 3D worlds with memory? That’s not just impressive, it's paradigm-shifting.

Because of artificial intelligence, our future has become much more interesting. We just think and we get the same thing created at the same time.