Support NeoGAF

[TheLikou]

Gaming? This shit can't generate good level design, it will just copy some things, steal from real creators. The only good thing that will really come out of this is porn.

 

[bighugeguns]

AI embracers and AI rejecters in here like...

They'll all fall in line to a man when their favorite game shows up pretty, quick and cheap. This is just useless noise from publishers trying to gaslight their own investors, or unions.

 

[HRK69]

I just don't know how to respond to any of you guys

That's not true. You've been handing out emotes like AI can't stop generating fingers

 

[StueyDuck]

It's pretty hard to wrap my head around what I'm seeing. I feel like one of those apes in Space Odyssey around the monolith. Don't really know htf it works but it seems incredibly important.

my assumption is its just running an AI filter like those ai game videos you see, except it is drawing many of those frames in real time. Which is impressive to do quickly.

but that one that literally does the exact "crazy trevor" animation seems like bull to me.

 

[Black_Mamba]

Interesting tech demo, but Take-Two down 10% because of this is crazy. We're closer to making movies with AI that will bankrupt Netflix than seeing this technology disrupting game companies. Both are still quite far, imo. I see a good trade opportunity today.

 

[bighugeguns]

That's not true. You've been handing out emotes like AI can't stop generating fingers

It's exciting tech, I'm allowed to be excited, I'm not going to call this stuff slop, sorry.

I think I'm entiteld to give reactions like everyone else here? Is something wrong?

 

[Zathalus]

I just don't know how to respond to any of you guys, it's over.

Maybe if you actually read my posts? Or asked what any of it meant? Instead of hopium that this technology is going to introduce some sort of gaming Nirvana?

Like the major one I mentioned. Temporal stability. Games are deterministic. You play a game, I play a game, it's the same game. Be it a decade between playthroughs. Current generative models struggle to replicate that on a frame by frame basis. Things can literally change completely if you simply keep playing. And not the same things, and certainly not in any sort of deterministic manner. Now early days with this technology, but it's still fundamentally a completely unsolved problem. And they don't really know how to fix it, among many other problems. No authoritative truth for one, which means a simple save/load that games have had since the 80s? Impossible.

So, feel free to hope away, but I'd wager by 2030 games will still be made the traditional way. Just with AI helping in code, voices, art, and such. Certainly not by creating games from a prompt. Hybrid integration is possible, but even then the complexity has a shit ton of unsolved problems.

 

[HRK69]

It's exciting tech, I'm allowed to be excited, I'm not going to call this stuff slop, sorry.

It's genuinely fascinating. That said, it's nowhere near the apocalyptic 'it's over' moment you were framing it as

Show me a 40-hour RPG with consistent systems, tone, and narrative before we panic.

 

[bighugeguns]

It's genuinely fascinating. That said, it's nowhere near the apocalyptic 'it's over' moment you were framing it as

Show me a 40-hour RPG with consistent systems, tone, and narrative before we panic.

You should probably panic.

Negativity drives my excitement. It means things are legit.

Coordinated negativity is going to excite the money too.

 

[Reckheim]

Nah, not this time, this shit is real. It's over. I want to play a new AAA deus ex in my lifetime.

I'm not saying AI wont get there, but this tech is basically just taking an image/video and making it interactive. You wont see games created with AI for a while.

it will be a good tool for prototyping tho.

 

[Fess]

I see your point, and you have many good ones. Consider the numbers, OpenAI lost $25 billion in 2025 alone, with another $14–$20 billion in expected losses for 2026. Even the paid tiers aren't helping much, hardware, energy, and research costs are much higher than people realize. You're right that they are gambling on growth for a future payout, but the scale of this risk is unique, which is exactly why there's so much talk of an "AI bubble" While OpenAI burns cash to lay the way, some competitors are simply sitting back on their own paid tiers, waiting for OpenAI to burn out so they can waltz in and capture a calmer market.

If OpenAI burn out then the AI industry is toast. That's what I think at least. But all of them need to find out a better way to pay for it all than what we have right now.

 

[ScHlAuChi]

That's not enough, I agree, but the other pieces are coming. For instance, rendering the world as something closer to Gaussian splats is very possible in the future, which already enable models to create entire 3d spaces that are navigable and photorealistic. We'll see hybrids of this soon enough.

Sure, and all that stuff is cool, but in terms of games, doing it the other way round makes much more sense.
So instead of trying to force this tech to do something it was never made for, it is much more logical to integrate this tech as a neural renderer in an engine, as then you get the consistency you need without the disadvantages of LLMs.
Like lets say you want to build a game like Bayonetta with complex combat, that would simply be impossible to describe in text.

 

[nemiroff]

AI embracers and AI rejecters in here like...

There was a fight / backlash against photography in the 1860s... It's hard for some peopl to look past their own noses when new tech is introduced. But on the other hadn, to be fair, there's a graveyard of "new tech" out there. 3D-TVs is a classic.

 

[IntentionalPun]

It's impressive, but hardly seems applicable to anything since it's not generating an actual game, but just endlessly generating video based on prompts/system prompts isn't it?

 

[falconhoof]

99% hype, 1% substance.

This has nothing to do with AAA dev, this simply wont EVER be ready for game development because of the tech it is based on!
Genie is an LLM based video generation system that makes things up as it goes, thats the nature of LLMs.
There is no underlying math, logic or physics that powers what you see on the screen.
What you see is just the "statistically" most likely output!


As a dev for over 22 years I can certainly say that we use alot of smoke and mirrors in games, but development is certainly not an illusion.
And if you think I´m scared of Genie - nope, I think it will become very useful as a local neural renderer for actual game engines in the future.
But that doesnt change the fact that Genie itself is basically a gimmick.


You are saying this in the exact same thread where people called those videos "gameplay" and are impressed by it ;P

^ Nice to read some sane posts in another thread of snake oil believers. Man it's like sitting around a bunch of blockchain sniffers back in 2018 telling me how it's all over and how their shit database was going to fix all the world's ills.

Blockchain is a tech and has a function.
AI is a collection of technologies and they have functions.

The function of this is not to replace game creation. It just does a kinda cool simple facsimile of it with a level of magical polish that fools the eye into thinking the fast food burger is gourmet.

There's no game engine, there's no data tracking, there's no actual physics. If you see a house, it's not an asset that can be recalled and rendered in the future, it's just a fuzzy set of weights in whatever temporal memory of the system has collected, which most likely will be forgotten. If you've ever had a conversation with ChatGPT and it suddenly forgets something you talked about earlier, or it invents something you didn't talk about, same thing here.

People are seeing a few second clips and suddenly thinking that they're going to be able to play some Skyrim like RPG where they can walk back to Whiterun after 50 hours and expect it to be the same. There's no technology in this that can do that stuff. It's not designed to and it's not one or two steps away from doing so; it's a completely different paradigm.

It's a wonderful academic project, and may have some utility in a niche space, but it ain't overtaking game development at all.

 

[ZoukGalaxy]

Summary of our (poor) future:

From the TV show WestWorld.

Just watch the 4 seasons, absolutely everything is there...

 

[Idleyes]

I'm just gonna mind my business and watch.

 

[ResurrectedContrarian]

So instead of trying to force this tech to do something it was never made for, it is much more logical to integrate this tech as a neural renderer in an engine, as then you get the consistency you need without the disadvantages of LLMs.

I agree that hybrids are the future. It's hard to say exactly where the line will be between the deterministic and sampled parts over time though.

 

[bighugeguns]

I'm just gonna mind my business and watch.

I like how there are people who think they're going to outfox Google.

 

[Zathalus]

I like how there are people who think they're going to outfox Google.

Killed by Google

Killed by Google is the open source list of dead Google products, services, and devices. It serves as a tribute and memorial of beloved services and products killed by Google.

killedbygoogle.com

 

[onQ123]

Dreams + AI

 

[MidGenRefresh]

^ Nice to read some sane posts in another thread of snake oil believers. Man it's like sitting around a bunch of blockchain sniffers back in 2018 telling me how it's all over and how their shit database was going to fix all the world's ills.

Blockchain is a tech and has a function.
AI is a collection of technologies and they have functions.

The function of this is not to replace game creation. It just does a kinda cool simple facsimile of it with a level of magical polish that fools the eye into thinking the fast food burger is gourmet.

There's no game engine, there's no data tracking, there's no actual physics. If you see a house, it's not an asset that can be recalled and rendered in the future, it's just a fuzzy set of weights in whatever temporal memory of the system has collected, which most likely will be forgotten. If you've ever had a conversation with ChatGPT and it suddenly forgets something you talked about earlier, or it invents something you didn't talk about, same thing here.

People are seeing a few second clips and suddenly thinking that they're going to be able to play some Skyrim like RPG where they can walk back to Whiterun after 50 hours and expect it to be the same. There's no technology in this that can do that stuff. It's not designed to and it's not one or two steps away from doing so; it's a completely different paradigm.

It's a wonderful academic project, and may have some utility in a niche space, but it ain't overtaking game development at all.

People who have no idea how LLM work are like this.

 

[HRK69]

You should probably panic.

Negativity drives my excitement. It means things are legit.

Coordinated negativity is going to excite the money too.

On a technical level alone, I find this AI work genuinely exciting

I'm honestly curious how this is supposed to be implemented in AAA development, given that Genie is a 60 second playground and not a game engine. Maybe my imagination and creativity are lacking here

 

[bighugeguns]

Did Google just kind of crash the games industry?

 

[nemiroff]

If OpenAI burn out then the AI industry is toast. That's what I think at least. But all of them need to find out a better way to pay for it all than what we have right now.

Yeah... the problem with AI right now, which results in a chaotic monetization, is that it's compressing what would normally be years of growth into months. And "AI slop" is just a reality of people misusing a tech that isn't finished yet. As a hobbyist javascript coder f.ex., using AI coding, I see "hallucinations" as a fascinating but very real danger that won't exist once the tech matures with persistent memory. And that's the real gamble: the billions being put in now are for the version of AI that actually works, not the "shambles" we sometimes see today.

...wait, I just realized I'm pretty much rambling at this point.

 

[Idleyes]

 

[bighugeguns]

On a technical level alone, I find this AI work genuinely exciting

I'm honestly curious how this is supposed to be implemented in AAA development, given that Genie is a 60 second playground and not a game engine. Maybe my imagination and creativity are lacking here

Tim Sweeney sounds pumped.



Looks like it can run doom.

 

[mopspear]

I saw this yesterday and thought it was funny. For context, it came out a month ago and was made with Sora-2.

 

[Idleyes]

Honestly, AI isn't going anywhere. Like most emerging tech waves, there will eventually be a correction when the focus shifts from hype to return on investment. Right now we're in the adoption phase, getting people to use it and build habits around it. When the bubble settles, only AI tools that are genuinely useful and sustainable will stick around. If Genie 3 eventually becomes able to handle game mechanics and systems, not just world generation, that could be a major shift, especially in the hands of experienced developers rather than someone like me giving it vague, generic instructions.

 

[SABRE220]

These are Gameplay, not video.

Isit though?... What hardware is it running on industry-class servers? Is it running in real-time? Will the data be stable and replicable in a format that can be stored and sold?
Explosive marketing can easily throw statements out that can be misinterpreted; hell, they are conveyed in a way to ensure that happens. Lets wait for more details to be revealed, Ai advancement is amazing but making complex game worlds in 5 minutes...probably not.

 

[bighugeguns]

Honestly, AI isn't going anywhere. Like most emerging tech waves, there will eventually be a correction when the focus shifts from hype to return on investment. Right now we're in the adoption phase, getting people to use it and build habits around it. When the bubble settles, only AI tools that are genuinely useful and sustainable will stick around. If Genie 3 eventually becomes able to handle game mechanics and systems, not just world generation, that could be a major shift, especially in the hands of experienced developers rather than someone like me giving it vague, generic instructions.

What's dawning on me is that it seems instead of looking at integration, publishers instead decided to run their media people to openly attack the idea of AI, alluding to lawsuits etc, nasty shit, makes sense looking at stocks right now, I guess. Maybe Google isn't interested in playing ball and to just rush to go straight for replacement. That's pretty fuggen metal.

Understand now why so many accounts here sound like pissed off teenagers.

 

[IntentionalPun]

Isit though?... What hardware is it running on industry-class servers? Is it running in real-time?
Explosive marketing can easily throw statements out that can be misinterpreted; hell, they are conveyed in a way to ensure that happens. Lets wait for more details to be revealed, Ai advancement is amazing but making complex game worlds in 5 minutes...probably not.

It runs in real time but it's all constantly rendering the video likely using far more energy / power than regular video generation would. And that's every second you are playing the game, since it's not producing a game executable to "play." Someone playing a "game" for a few hours would probably cost literally thousands of dollars lol

 

[bighugeguns]

It runs in real time but it's all constantly rendering the video likely using far more energy / power than regular video generation would. And that's every second you are playing the game, since it's not producing a game executable to "play." Someone playing a "game" for a few hours would probably cost literally thousands of dollars lol

There absolutely would be a way to slice off boundaries and export it to an executable so you can share your games with your friends.

Running counter narratives against this tech is fucking sad and a waste of money.

 

[SABRE220]

It runs in real time but it's all constantly rendering the video likely using far more energy / power than regular video generation would. And that's every second you are playing the game, since it's not producing a game executable to "play." Someone playing a "game" for a few hours would probably cost literally thousands of dollars lol

So, it isn't running in real time on local hardware, which is what I expected. This is essentially like the xbox running on the cloud power marketing scenario, but with some heavy parameters flipped, and the question would also comeup how to make a game that is consistent, replicable across all users, especially since the Ai platforms have trouble with memory and making mistakes.

 

[MidGenRefresh]

People who think this is running locally clearly haven't tried running LLM models locally.

 

[IntentionalPun]

There absolutely would be a way to slice off boundaries and export it to an executable so you can share your games with your friends.

Running counter narratives against this tech is fucking sad and a waste of money.

I'm just explaining what it's doing not running a counter narrative.

 

[bighugeguns]

I'm just explaining what it's doing not running a counter narrative.

Oh sorry, not directed at you, just the vibe that's weird.

 

[Idleyes]

There absolutely would be a way to slice off boundaries and export it to an executable so you can share your games with your friends.

Running counter narratives against this tech is fucking sad and a waste of money.

That doesn't exist yet, but the idea itself makes sense. What you're describing would be like an AI engine that can take what a generative world model produces and translate it into the structured data and systems that engines like Unity or Unreal actually require.

 

[viveks86]

I find it hard to believe the people creating actual science fiction shit won't be able to port this stuff to pre rendered dynamic engines.

You'll need a technical artist to cut out the boundaries for render etc.

I really look forward to what professional game devs does with this. The speed of development alone will be awesome.

Sorry to say, but you are getting wayyy too carried away by this.

NONE of this is happening using Genie 3 or any such world model.

There is no boundary to cut out and render. There are no 3d models or meshes. There are no actual particles or objects. No physics calculations or lighting models. There is no actual NPC routines, path finding or simulations. What it's doing is predicting frames based on a scenario, keyboard input and constraints. All that exists are these predicted 2D frames, reaction to user input and some memory to maintain consistency across frames.

This is not how any real game should be built as every frame is a hallucination constrained by the previous frame and the original prompt. Minutes later, all is forgotten. There is no permanence other than recorded footage. You quit and restart, and it's not even the same thing. There is no understanding of the world. That mini-map in "GTA" for example, is also a hallucination that may or may not have exact correlation to the "world" being generated. It's just pretending based on what it thinks a mini-map would look like based on the previous frame and keyboard interaction. Look at those street lights where both red and green are glowing. There is no true understanding of how the real world works. It's just dreaming it up as it goes, based on all the shit on the internet that ever existed, including copyrighted material. It's pretty damn cool for academic reasons. But people betting on stocks based on it? Fuckin' LOL.

Generative AI could certainly help in automating a lot of grunt work on the development side, like building textures, voice generation etc. Down the line, I'm sure usable base 3d models will be generated offline too to save dev time that a modeler can iterate on. And it can do a lot on the final rendering side that we are merely scratching the surface on. Read up on recent papers on the topic of (non-generative) AI/ML based lighting models that eliminate the need for ray tracing, such as Lightformer and NeLiF. Still early days, but THAT's the future for AI in gaming. And at some point LLMs will be used to generate NPC dialog/allow free form conversations and make contextual decisions in game etc. But this is none of that. It's great for conceptualization and messing around for fun. It'll never help with productivity.

 

[Cancer]

Music / muzak creation has been around for quite a while now and is accessible and of pretty high quality.
I rest my case.

 

[simpatico]

I've been kicking around the same idea for a game for the better part of a decade. You better believe I'm making it in 2030 with this technology. You guys are gonna be happy I did, promise.

 

[bighugeguns]

That doesn't exist yet, but the idea itself makes sense. What you're describing would be like an AI engine that can take what a generative world model produces and translate it into the structured data and systems that engines like Unity or Unreal actually require.

Ye, concatenate the scene for export, add UIX afterwards etc, in terms of tooling in a process this makes sense, though I don't know what Google's goal is, judging by the complete bitchfits being broadcast by publisher partners I guess they're being sidelined here.

 

[IntentionalPun]

Oh sorry, not directed at you, just the vibe that's weird.

Well your suggestion is still going to require some totally different technology than what Genie is doing here.

I don't think it's particularly fair to act like people who understand this are running some sort of counter narrative.

 

[bighugeguns]

Well your suggestion is still going to require some totally different technology than what Genie is doing here.

I don't think it's particularly fair to act like people who understand this are running some sort of counter narrative.

You're right, so there's no way to integrate this into current games development? Is that why money is moving away from publishers? I thought the opposite would happen.

 

[chilichote]

Great for VR if you want to spontaneously escape to a place of your choosing. But far too sluggish for gaming.

 

[Idleyes]

This is where we are now.

 

[IbizaPocholo]

 

[neocycle]

These are not games. Walking simulators are more fun than this crap.

 

[kiphalfton]

Lmao, there are so many copyright being broken by all those, ifs meaningless.

 

[IntentionalPun]

You're right, so there's no way to integrate this into current games development? Is that why money is moving away from publishers? I thought the opposite would happen.

I really doubt investors grasp the technology limitations, they just see something resembling progress towards "replacing traditional video games" and are dumping stocks.

IMO way too premature; it's far more likely that AI tools will empower game developers not replace them, and that should not cause stocks to drop. Those tools already exist really and are improving every day. But this really isn't that.