Research at NVIDIA: The First Interactive AI Rendered Virtual World

[Alucardx23]

This is the type of thing that gets me really exited for the near future. This will really help to improve the graphics quality and lower the production cost and development time for games.

Nvidia has created the first video game demo using AI-generated graphics

"It's a new way to render video content using deep learning," Nvidia's vice president of applied deep learning, Bryan Catanzaro, told The Verge. "Obviously Nvidia cares a lot about generating graphics [and] we're thinking about how AI is going to revolutionize the field."

"The structure of the world is being created traditionally," explains Catanzaro, "the only thing the AI generates is the graphics." He adds that the demo itself is basic, and was put together by a single engineer. "It's proof-of-concept rather than a game that's fun to play."

https://www.theverge.com/2018/12/3/...deo-game-graphics-nvidia-driving-demo-neurips

NVIDIA Invents AI Interactive Graphics

"Twenty-five years ago NVIDIA transformed the computer graphics industry by building the first GPU, the modern day tool of the da Vincis and Einsteins of our time. Now, a new deep learning-based model developed by NVIDIA researchers aims to catapult the graphics industry into the AI chapter.

Using a conditional generative neural network as a starting point, the team trained a neural network to render new 3D environments, after being trained on existing videos.

This AI breakthrough will allow developers and artists to create new interactive 3D virtual worlds for automotive, gaming or virtual reality by training models on videos from the real world. This will lower the cost and time it takes to create virtual worlds.".....

https://news.developer.nvidia.com/nvidia-invents-ai-interactive-graphics/?ncid=so-you-ndrhrhn1-66582

 

[modiz]

ummm this is pretty crazy.... and kind of weird to think that in the not so far future the AI will be able to generate video game levels.

 

[Soapbox Killer]

Damn future, you scary!

 

[Arkanius]

*Bethesda wants to know your location*

 

[Sabretooth]

That Gangnam Style video at the end is nuts. It kind of reminds me of Deepfakes.

 

[Arkanius]

That Gangnam Style video at the end is nuts. It kind of reminds me of Deepfakes.

Click to shrink...

Practically the same principle

 

[Femto0]

Skynet is close

 

[Kemono]

This stuff has extreme potential.

Sometime in our future we'll be able to tell an ai what kind of game we want and it can generate it for us.

Futurama had robot writers getting oscars in the year 3000 but i think we don't have to wait that long.

 

[lazerfox]

Using machine learning so you don't have to embarrass yourself dancing to gangnam style. Genius!

 

[Alucardx23]

This stuff has extreme potential.

Sometime in our future we'll be able to tell an ai what kind of game we want and it can generate it for us.

Futurama had robot writers getting oscars in the year 3000 but i think we don't have to wait that long.

Click to shrink...

I was thinking exactly that. Games made just for you. 10 out of 10 games for everyone! :D

Check this video out so you can see where we are at this moment on that.

 

[brainchild]

This shit is starting to get really creepy...beam me up, Scotty! 😂

 

[WellbornWarrior]

That's very interesting but doesn't that technology have the drawback that it only works with real world objects or locations? I mean you can't make a fantasy or Sci-Fi environment with that.

 

[Vanta Aurelius]

So is the the AI creating the textures as well, or would artists still need to do that? Amazing technology regardless!

 

[Alucardx23]

That's very interesting but doesn't that technology have the drawback that it only works with real world objects or locations? I mean you can't make a fantasy or Sci-Fi environment with that.

Click to shrink...

We will eventually be able to train Neural networks with all of the fantasy or Sci-Fi content that is out there (Pictures, Videos, Books) and then create original content based on that. This is not something that is working perfectly right now, but the basics are there already for the technology to grow on.

 

[Deleted member 49319]

So this is deep dream in 3D?

 

[Railgun]

This is really exciting stuff

 

[iksenpets]

I love how this video is like check out this tech we could use for procedural videogame environments, and then as a little aside at the end, also throw in how they could use it to forge evidence of anyone performing any action

 

[Coconico]

Man, even the guys who make the computers are gonna be out of a job when AI becomes the labor standard! We're very quickly hitting the point of no return, I was hoping I wouldn't see it in my lifetime, but here we are.

 

[DarrenM]

Sometime in our future we'll be able to tell an ai what kind of game we want and it can generate it for us.

Click to shrink...

One step closer to the Holodeck

 

[Deleted member 4093]

That Gangnam Style video at the end is nuts. It kind of reminds me of Deepfakes.

Click to shrink...

Reminds me of a snapchat filter

 

[Deleted member 4761]

 

[Alucardx23]

So is the the AI creating the textures as well, or would artists still need to do that? Amazing technology regardless!

Click to shrink...

On this example the Neural Network was trained using thousands of videos from the Cityscapes, the content is segmentation based on type (street, car, person, tree, building, etc etc) once it has that information classified it is able to build and populated a city. So to answer you question, the textures are generated based on what it learned from the real word and I understand they can modify them as well.

 

[Deleted member 13645]

Wow, this is fascinating. Nvidia is doing some really interesting work. I'm excited to see where this goes.

 

[FreDre]

This is nuts.

You could use it to replicate entire cities by using this deep learning technique with a database of procedural 3D models, such as trees, or cars.
They could pair it with a website that offers those.

This is going to cut down a lot of the budget for mid-sized games.

 

[Dekuman]

I'd like to live there.

 

[Dynamite Cop]

Did they flip that video of the dancing girl and cut off the url so they don't have to credit her? lol

 

[gozu]

Doesn't look anywhere near prime time. The "deepfake-style" video of his colleague is worse than most deepfakes as well.

I am not impressed right now. Yeah yeah potential bla bla bla. Wake me up for v3.0.

 

[Kadath]

Replicating isn't the same as creating. This isn't so unlike generating levels in Diablo or No Man's Sky. We are already doing that.

Teaching an AI game design... is a different problem.

Edit: Oh, and the article actually says the 3d lanscape is done manually as usual. The only thing AI is doing is generating and drawing textures on top of it.

I'm pretty sure we can stick to artist-created textures for a while, still. (or you'd have to scrap all lighting/raytracing whatever is used for illumination, essentially this AI stuff overlaps/replaces everything else)

 

[FreDre]

Doesn't look anywhere near prime time. The "deepfake-style" video of his colleague is worse than most deepfakes as well.

I am not impressed right now. Yeah yeah potential bla bla bla. Wake me up for v3.0.

Click to shrink...

It's a research.
Of course it's not ready.

 

[Deleted member 20284]

So when the robots wipe one of us out they can simply install an imposter as long as they wish.

Here is us taking a selfie with Ozzy, video of him dancing, he's a good dancer...see he's not dead at all. Beep Boop. Move along citizen.

 

[Hamlet Machine]

The secret services are gonna love the future…

 

[gozu]

It's a research.
Of course it's not ready.

Click to shrink...

Nowhere near primetime -> somewhat near prime -> near prime time -> ready -> standard.

Hope this clarifies things for you. I am a stickler about precise language.

 

[RestEerie]

Nowhere near primetime -> somewhat near prime -> near prime time -> ready -> standard.

Hope this clarifies things for you. I am a stickler about precise language.

Click to shrink...

 

[gozu]

Click to shrink...

What can I say. I was excited by the title. i watched video. I was disappointed. I shared my feelings, while trying to imply the tech still had promise for the future

Want to have fun? Go play RDR2. It's awesome! Dutch is an amazing character. Like goddamn!

 

[BreakAtmo]

Awesome. The beauty of it is that the graphics the AI generates don't even necessarily have to be amazing. Even when devs will have to clean up the visuals later it'll still save huge amounts of time, effort, and money. And if the AI can monitor those development processes, it'll likely improve even further.

 

[Powerofthecloud]

Nowhere near primetime -> somewhat near prime -> near prime time -> ready -> standard.

Hope this clarifies things for you. I am a stickler about precise language.

Click to shrink...

Yes, prime time. A precise and commonly used word by all scientist around the world to describe readiness.

 

[Deleted member 7130]

When will they render electric sheep?

 

[Instro]

I wonder how useful this could be to unique worlds, rather than ones based in reality. I'd hate to see even more modern and contemporary settings for games just because AI becomes a useful dev tool.

 

[gozu]

Yes, prime time. A precise and commonly used word by all scientist around the world to describe readiness.

Click to shrink...

Thank you for saving me from myself. I was wrong, I *AM* excited.

Are you happy now?

 

[Alucardx23]

This one shows how they can edit and configure a scene. Pretty awesome stuff in my opinion and really shows how flexible this technology will be.

 

[Alucardx23]

The technique is now a lot more consistent between one frame and another.

 

[brawndolicious]

The technique is now a lot more consistent between one frame and another.

Click to shrink...

Using AI to generate a second viewpoint can be a major leap for VR gaming. Assuming it's less intensive than traditional rendering, you won't need to double the processing load to create an image for each eye.

 

[Alucardx23]

Using AI to generate a second viewpoint can be a major leap for VR gaming. Assuming it's less intensive than traditional rendering, you won't need to double the processing load to create an image for each eye.

Click to shrink...

Yes, for VR this should eventually be a big performance boost. I love how Nvidia is really pushing machine learning for gaming.

 

[Deleted member 76797]

Nvidia employs some really brilliant people and it's paying dividends