NVIDIA’s New AI: Virtual Worlds From Nothing! + Gemini Update!

📝 The papers are available here:

📝 My latest paper on simulations that look almost like reality is available for free here:

Or this is the orig. Nature Physics link with clickable citations:

🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
Alex Balfanz, Alex Haro, B Shang, Benji Rabhan, Bret Brizzee, Gaston Ingaramo, Gordon Child, Jace O'Brien, Jie Yu, John Le, Kyle Davis, Lukas Biewald, Martin, Michael Albrecht, Michael Tedder, Owen Skarpness, Richard Putra Iskandar, Richard Sundvall, Taras Bobrovytsky, Ted Johnson, Thomas Krcmar, Tybie Fitzhugh, Ueli Gallizzi.

#nvidia #gemini

[virtual-reality]

amazing videos!
you have inspired me to combine 3D gaussian splatting for realtime radiance gield rendering with medical imaging

[fireguy]

Exponential growth in these fields are so mind boggling

[RaumBances]

No worries about Google Gemini. You present relevant and timely material that you accept from sources in good faith. Slight of hand by Google is not your fault. I appreciate the follow up and completely agree that they should have opened that video with a disclaimer and brief description of how the video was put together for presentation. It may have initially impressed people but in the long run, it disappointed most because no one appreciates being tricked.

[Neo_mancer]

What a time to be alive is the most iconic line I have ever heard

[GomesBrown]

What a time to be alive! As an artist it's indeed a miracle I'm still alive.

[IceMetalPunk]

I almost want to build a pipeline of AIs where you describe a character doing an action, an LLM separates the character from the action, text-to-3D generates the character in a T-Pose, an autorig technique rigs it up a la Mixamo, and then one of those text-to-animation models animates it from the action you described. So something like "a wizard doing jumping jacks" just works...

[somethingtojenga]

They're going backwards out of the scene because it's easier to create information around a frame and have existing information get smaller and more obscured as the scene progresses. It's comparing apples to oranges, an outdoor scene flying toward a small part of the horizon and having to make it bigger and more detailed, vs. making things up and having them get smaller and less detailed.

[hbsharkman]

Cool that you followed up with the Gemini correction.

[hanktremain]

that shadow removal will be super useful for 3D texturing

[TMGreycoat]

That PatchMatch video was the first of yours I've watched. Amazing to see how far things have come

[marcfruchtman]

Video to Mesh... that is really incredible.

[xingzhexin]

Whoa! Shoutout from Jim Keller! Dr. you are awesome and it's great to see more recognition!

[viddarkking]

This papers are awesome but to be truly useful at an efficient level we need the tools and UIs to make proper use of them.
For example With the text to 3D If I have a character and want to create alternate costumes I need to be able to use the "Character Base Mesh" as a guide for the AI.

[jk_lol]

thank you for correcting the record on gemini. they tried to pull the wool over our eyes.

[kinngrimm]

1:10 I would guess it needs some stability, some constants that would allow it to continue its extrapolations. If one would use this not as a video, but a gaming background instead within which we as players would be part of the scene, maybe that would do the trick. From that then our actions would narrow down the realm of possibilities into which the extrapolations could exapnd into.

[nicolaastanghe]

omg the shadow removal :O I wonder if it`s good enough that it can keep landmarks consistant when multiple photos are taken from multiple angles.

[baslifico]

Congrats on the recognition dude. You've earnt it.

[espace]

FunSearch: Making new discoveries in mathematical sciences using Large Language Models

[hazelhulet]

Really impressed with your scolarly integrity on Gemini. It was so important to get this correction out.

[bigmouthfisheyes]

Very dreamy. The way everything is incoherent and kind of melts together. All of it is morphing constantly.