Google's New Genie 3 AI Model Creates Video Game Worlds In Real Time (theverge.com) 15
An anonymous reader quotes a report from The Verge: Google DeepMind is releasing a new version of its AI "world" model, called Genie 3, capable of generating 3D environments that users and AI agents can interact with in real time. The company is also promising that users will be able to interact with the worlds for much longer than before and that the model will actually remember where things are when you look away from them. [...] Genie 3 seems like it could be a notable step forward. Users will be able to generate worlds with a prompt that supports a "few" minutes of continuous interaction, which is up from the 10-20 seconds of interaction possible with Genie 2, according to a blog post.
Google says that Genie 3 can keep spaces in visual memory for about a minute, meaning that if you turn away from something in a world and then turn back to it, things like paint on a wall or writing on a chalkboard will be in the same place. The worlds will also have a 720p resolution and run at 24fps. DeepMind is adding what it calls "promptable world events" into Genie 3, too. Using a prompt, you'll be able to do things like change weather conditions in a world or add new characters. The model is launching as "a limited research preview" available to "a small cohort of academics and creators," according to Google. It's "exploring" how to bring Genie 3 to "additional testers."
Google says that Genie 3 can keep spaces in visual memory for about a minute, meaning that if you turn away from something in a world and then turn back to it, things like paint on a wall or writing on a chalkboard will be in the same place. The worlds will also have a 720p resolution and run at 24fps. DeepMind is adding what it calls "promptable world events" into Genie 3, too. Using a prompt, you'll be able to do things like change weather conditions in a world or add new characters. The model is launching as "a limited research preview" available to "a small cohort of academics and creators," according to Google. It's "exploring" how to bring Genie 3 to "additional testers."
They think they're inventing the holodeck (Score:2)
But really, they're just boiling oceans for a poor implementation of Rogue.
Re: (Score:1)
Have you seen modern games? I'd be happy for a "poor implementation of Rogue" at this point.
Tarkov is now one of the best games on the market, how far we've fallen.
Tho I agree with you, this is gonna suck.
The tech is fascinating (Score:2, Interesting)
It looks fascinating but surely this tech cannot possibly be a substitute for actual worldbuilding and writing out detailed game logic. Lack of real persistence kills any serious applicability... unless a game is deliberately designed to lack persistence, perhaps simulating a dream.
Could be useful in game design though, for inspiring creators and quickly testing out ideas.
Re: (Score:2)
I think the question to ask is if gamers are looking for art, for some aesthetic and comment on the human condition. Or are gamers here to be entertained, to consume content in any way that satisfies the immediate desire for stimulus.
Re: (Score:2)
Using AI for what its bad at (Score:2)
The other day someone mentioned that most AI art sucks because it's unskilled people trying to get AI to do shit its bad at. They were referring to how interesting deep dream was compared to most of what gets made these days. It's a very interesting point. I'd really like something like this for a LSD dream emulator, i think doing something in VR and maybe even having an algorithm that uses eye tracking to discern what catches the wearer's eye.
I know it'd never be a major application, but trippy art is d
This is cool but how much $? (Score:2)
Their prentation was cool to watch but it didn't mention the number of video cards that it takes to run and the power bill. I would love to know what those numbers are.
It would be very cool if demand for this kind of ultra-complex simulation leads to amazing hardware innovation - but at the moment its looking like its just about buying vast numbers of video cards instead of building better ones. Maybe the market will eventually create really great new hardware based on this demand that can cheaply run amazi
Re: This is cool but how much $? (Score:2)
And here I am wondering how Google is even going to monetize their models. Are they going to shove ads into their models so every time you prompt something the model will subtly generate some form of product placement? Google has always struggled with the business aspect of their tech outside of the advertisement model.
Actually that goes for almost every AI company out there. Subscriptions alone will not cut it as eventually you hit a demand ceiling where everyone who wanted your service will have already p
Likely will not feel like a place (Score:3)
Obvious bullshit (Score:2)
At least to any gamer. Generated game-worlds are universally boring and non-engaging. The engagement comes from actual game designers placing stories in the world. And that is the part that takes real work and real creativity.
Re: (Score:2)
You think that is what they're going for? They don't care about games. This is about advancing the field of AI in general with an incredible improvement over what was previously possible (and synthesizing training data for robots). It is an incredible achievement made possible with tons of R&D money and intelligence being thrown at it.
Stop being such a cynical hater.
Wasted effort (Score:2)
Is this the right tool for the job? Procedural environment generation has been around for awhile. Many games do this.
Re: (Score:2)
Is this the right tool for the job? Procedural environment generation has been around for awhile. Many games do this.
Yes, but you can't sell procedural environment generation as AI, and as we all know, AI makes everything better. "But it uses more energy for a less viable end-product." Silence! AI IS BETTER!
Cool demo (Score:3)
And can even have some use cases. But in videogames? I don't think so. It can be interesting for concept, but if I add a dragon, I don't want it to ignore everything, I want some fight mechanics making the game interesting, and if I have to set them on the prompt, I can better program them in LUA. This will burn an immense amount of resources just to generate an uninteresting environment in 720p@24FPS that will only last "for a few minutes". You need a lot more for a videogame.