• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

After Sora by OpenAi, AAA games by AI are months away

cash_longfellow

Gold Member
The insane amount of “I don’t think AI could do this” here in this thread is insane to me. You know what a bunch of other human beings didn’t think could happen outside of a few geniuses over the last century? Windows, Video Game Consoles, TVs, Air Planes,
I wonder what those coal miners are going to do. Right after they are told “learn to code”, they get comfy at their new job coding and bam, they get done in by AI. They are going to need another new career field.
“career field” is your biggest concern with this topic?? That’s cute 😂
 

ProtoByte

Member
Its going to be harder and harder to actually have a skill that makes you fucking money. AI just going to replace everything. Fucking ridiculous. You think its hard to make a living now, wait til AI eliminates every goddamn fucking job/hustle on the planet. I will not cheer about this.
AI requires a lot of human interaction, and not all jobs can even incorporate AI. I wouldn't get so worried about it. Prepare, yes. Retrain. Dooming about it won't help anyone.
 

kruis

Exposing the sinister cartel of retailers who allow companies to pay for advertising space.
This thread is full of people who think an AI generated video of Minecraft is the same as Minecraft the 3d game running on your PC. The two are not alike.

If Ai is going to change game development, you’ll need AI tools that for example generate complete 3D models from text input and ready to use in Unreal Engine. A video of that 3D model is useless.
 

WakeTheWolf

Member
WbOAweP.gif


Edit - also, good luck if you are looking for work and didn’t create the perfectly tailored resume to the position…you won’t even be considered for an interview even if you are the most qualified…AI is a fucking joke and if you support it now, good luck when you legitimately need to talk to a human being in about 10 years for literally anything from customer service to food ordering. It’s sad that we aren’t even standing up against this nonsense right now….you, me, all of us will seriously regret it.
Not talking to other humans is a blessing.
 

CLW

Member
I'm not generally a proponent of AI - I an a "specist" - but there is ZERO doubt this is coming and it will 100% decrease labor costs causing mass layoffs of people which will lower the extremely too large budgets for the AAA gaming space. Of course the "no crunch" FKA "overtime" crowd will protest/strike making it really easy for the suits to decide who is easily expendable.
 

cash_longfellow

Gold Member
Not talking to other humans is a blessing.
Indeed! But I do still love talking to Gaffers. Just because I know we are all stupid and share some nutty feels about video games. I really do hope the idea of random feelings about video games (or really any feelings at all) doesn’t end!. I wouldn’t hold my breath tho 🤷‍♂️
 

kungfuian

Member
This thread is full of people who think an AI generated video of Minecraft is the same as Minecraft the 3d game running on your PC. The two are not alike.

If Ai is going to change game development, you’ll need AI tools that for example generate complete 3D models from text input and ready to use in Unreal Engine. A video of that 3D model is useless.
Not to dismiss the differences in complexity, because what you are saying is a valid point, but you have to think about the time frames and trajectory of the whole thing.

We've gone from translating text to high quality image generation all the way to text to high quality video, with realistic physics, basically over night. I don't think it's a far stretch to think the next evolution of this tech is text to the creation of full 3d worlds with rule sets, objectives, etc. I.E. games.
 

T-0800

Member
WbOAweP.gif


Edit - also, good luck if you are looking for work and didn’t create the perfectly tailored resume to the position…you won’t even be considered for an interview even if you are the most qualified…AI is a fucking joke and if you support it now, good luck when you legitimately need to talk to a human being in about 10 years for literally anything from customer service to food ordering. It’s sad that we aren’t even standing up against this nonsense right now….you, me, all of us will seriously regret it.
Everything will be fine.
 

Buggy Loop

Member
Hell, the first steps will probably be a mix like some very basic "prototype" esque looking game, where devs will at the very least define the rules, gameplay, limits, etc

It could look like this in the render stage

images


and the AI transforms it to


Artists will be first to lose their jobs with this. There will be a lot of complaints and it's gonna suck for a ton of peoples, but this is inevitable. It's like trying to have empathy for the monks who lost their medieval manuscript jobs when we invented the first printer.

Whole industries will disappear.... Also new ones will be created.
 
Last edited:

reinking

Gold Member
Part of the reason I like video games is that I appreciate the effort that has been put into creating art and crafting stories. If that goes away, I am honestly not sure how I am going to feel toward the medium. The same goes for movies. It is a fascinating tech, and I believe it can do a lot of good, but man... ...this is an insane amount of advancement in a short period of time. It will not stop at arts & entertainment. I can see a future where we will have AI and automated bots creating and mass producing material goods.

Family Guy Reaction GIF
 
Last edited:

WakeTheWolf

Member
What a stupid take.

Have you ever tried customer service with some stupid chat bot? Especially over the phone, no matter how advance it gets - Give me the HUMAN.
You clearly never dealt with Amazon customer service reps kek. Nice optimism I guess but I'll take the more adequate AI over the moron lazy human any day.
 

Buggy Loop

Member
The opposite is true, this is the main area they are struggling with, the white paper makes it clear that whilst they've now mostly solved persistence, physics and real world interactions are an abstract problem that's hard to teach a visually focused AI.

discussion_0.mp4

Yea but look how it improved so fast.

Look at the reflections on that video! AI is doing better than path tracing, outside of the visual bug, this is life-like. It'll be faster to fix the bugs than it will be for traditional methods to render that quality in games.

We have peoples working years on games and they have glitches
 
Last edited:

Reaseru

Member
"It's stealing from other videos...Stutterstock, blah, blah, blah."

Of course it has to learn from other videos. No entity learns to create something from nothing. It needs some foundations to learn.

We speak because we learned to join words together. Words that we learned by listening others. But the content of our own speech is of our own creation. It expresses what we think.
 
Last edited:

odhiex

Member
Sure it could very well be, with NVIDIA and Epic as among the major players to push the A.I. adoption.

I just wish game creators could use these technologies wisely and not too dependant on them.
 

Buggy Loop

Member
Obligatory two minute papers (I truely believes this guy put his words into a soundpad and AI is making his voice-over videos :messenger_tears_of_joy: )



This is the biggest advancement in AI, so far.
 

SF Kosmo

Al Jazeera Special Reporter
In an implementation like Cyberpunk with Overdrive RT, DLSS3 and Ray reconstruction, about 80% of the pixels you see are generated by AI. That number is only going to go up.

It's not going to be like some text prompt generation thing; in fact, I think AI is going to shift increasingly away from that paradigm toward models that allow more artist control and authorship. AI is going to be more useful when it speeds up the work of artists rather than attempting to poorly gronk what the user wants from a vague description.
 

Audiophile

Gold Member
My hope is that AI can be used to not just build outwards but to build inwards; and assist devs in doing things they never could have done before. Like making most interiors in the likes of GTA not just enterable but tying them into game systems and the overall simulation so that you feel like there's no barriers as you peer into the world. In addition, they'd appear highly variable by using morphed and modular assets with a vast array of style variables.

This is a slightly different realm but I envision it as devs building a building 'shell' as they normally would. Then tagging that building as a certain type or subtype (hotel, home, office etc) that an ai-procedural toolset will recognise. For each of these types the devs/artists will provide a set of example layouts/designs from which the system will learn. Said system will be aware of windows, doors etc. and any bespoke transitional spaces. The ai-procedural system will then fill out these buildings according to the types, using the artist examples; making it artist-driven. Over time certain layouts/variations could be rated/approved/disapproved and in turn the system will learn; in turn becoming better. Further embellishment could then be done on near-final results to varying degrees; these embellishments could themselves feed back into the system. This would effectively run as a virtual conveyor belt to be ok'd and qc'd during dev. Each room would have a seed for persistence.

Once everything is constructed, various gameplay systems could be tied into these spaces with the tagged "types" influencing NPC behaviour within them as well.

I don't wanna shake off the cops in GTA by going round the block and hiding any more. I wanna enter a hotel, take the lift and go haul myself up in a room, during which, as I enter all these nooks and crannies, I wanna feel like I'm peeling back the curtain on the lives of NPCs: before I enter said room I knock on the door and show a fake badge (other options include picking it and blowing it open), they let me in and I hold the guy up, his wife was sat on the sofa watching TV but now she's cowering in the corner. I bash a door open on the way through and some guy is just sat there taking a dump, he runs waddling into the living room corner with his pants round his ankles and the guy I'm holding up joins him. I proceed to steal their stuff to sell later. Then, a SWAT team comes to hunt me down, but I've stuck C4 in the hallway and I'll blow 'em up. Then I'll make my way to the roof, jump off and open my chute, land on a boat in the river, steal it and make my way back to the safehouse.
 
Last edited:

Dutchy

Member
This thread is full of people who think an AI generated video of Minecraft is the same as Minecraft the 3d game running on your PC. The two are not alike.

If Ai is going to change game development, you’ll need AI tools that for example generate complete 3D models from text input and ready to use in Unreal Engine. A video of that 3D model is useless.

This. It's actually insane how much people are misinterpreting how this works and what is actually necessary to create a seamless gaming experience. It can generate all the video it wants. If you want AI to build an entire game from scratch then you would need access to the raw data. Which would also have to be easily modifiable and compatible with various other pieces of software. But this would make it open-source beyond any reasonable comprehension and I honestly don't know when/if this is going to happen.

Not saying it's never going to be possible but saying we're only months away like OP is implying is utterly ridiculous.
 
The opposite is true, this is the main area they are struggling with, the white paper makes it clear that whilst they've now mostly solved persistence, physics and real world interactions are an abstract problem that's hard to teach a visually focused AI.

discussion_0.mp4



Simple games are based off simple examples of which many exist on the web for it to learn from, the more specific or elaborate you get, the worse the results become. Also all GPT models have a token length limitation, they can't "think" larger than a certain amount of code, so couldn't work on large projects without human intervention to piece things together.
I mean from where we were a year ago to now? I definitely underestimated AI’s ability to have consistent characters - it seems like they’re making really fast progress. What exactly do you think are the diminishing returns it will run into? I’m surprised you doubt the technology to this degree
 

I_D

Member
I think a lot of people are thinking slightly too far ahead, and therefore claiming it to be unrealistic.
If you're expecting everything to be generated in real-time, then yes; it'll be a while before hardware is good enough to do such things.

However, if you, instead, look at the baby-steps...

I really don't think it will be long from now when we see Borderlands/Remnant/Bloodborne Challice Dungeon/Phantasy Star Online/NIOH/etc. - style games' levels designed/generated by an AI.

The type of game in which grinding/looting is a core mechanic would benefit immensely from an AI-generated level-designer.
Considering that high-level developers have already had similar systems for more than a decade, I expect AI will allow just about anybody to do the same thing within a year or two, or so.

By this time, five years from now, I would not be even remotely surprised if tons of games had never-the-same questlines, where AI generates a brand new storyline every time you do something different.

It might not be done in real-time - but it might not even matter?
 

PotatoBoy

Member
Lol at the current cost they're paying to run these models, maybe they'll let you play an AI-generated game in a museum for $50 an hour.
 

Buggy Loop

Member
It's not just an AI generated video of minecraft. Again, from the papers

Sora is also able to simulate artificial processes–one example is video games. Sora can simultaneously control the player in Minecraft with a basic policy while also rendering the world and its dynamics in high fidelity. These capabilities can be elicited zero-shot by prompting Sora with captions mentioning “Minecraft.”

It's generating a world and understands it. This is no longer the dreaming god AI.

Nvidia's AI expert on the subject



Apparently some folks don't get "data-driven physics engine", so let me clarify. Sora is an end-to-end, diffusion transformer model. It inputs text/image and outputs video pixels directly. Sora learns a physics engine implicitly in the neural parameters by gradient descent through massive amounts of videos.

Sora is a learnable simulator, or "world model". Of course it does not call UE5 explicitly in the loop, but it's possible that UE5-generated (text, video) pairs are added as synthetic data to the training set.

If you think OpenAI Sora is a creative toy like DALLE, ... think again. Sora is a data-driven physics engine. It is a simulation of many worlds, real or fantastical. The simulator learns intricate rendering, "intuitive" physics, long-horizon reasoning, and semantic grounding, all by some denoising and gradient maths.

I won't be surprised if Sora is trained on lots of synthetic data using Unreal Engine 5. It has to be!

Let's breakdown the following video. Prompt: "Photorealistic closeup video of two pirate ships battling each other as they sail inside a cup of coffee."

- The simulator instantiates two exquisite 3D assets: pirate ships with different decorations. Sora has to solve text-to-3D implicitly in its latent space.
- The 3D objects are consistently animated as they sail and avoid each other's paths.
- Fluid dynamics of the coffee, even the foams that form around the ships. Fluid simulation is an entire sub-field of computer graphics, which traditionally requires very complex algorithms and equations.
- Photorealism, almost like rendering with raytracing.
- The simulator takes into account the small size of the cup compared to oceans, and applies tilt-shift photography to give a "minuscule" vibe.
- The semantics of the scene does not exist in the real world, but the engine still implements the correct physical rules that we expect.

Next up: add more modalities and conditioning, then we have a full data-driven UE that will replace all the hand-engineered graphics pipelines.

But what does he knows

This AI with spatial awareness and how things should not collide, understanding of physics and how it should react is massive and goes even beyond videos and games, this is something self-driving cars have completely lacked. This will also boost autonomous cars / robots.
 

Hoppa

Member
I'm not a huge fan of AI but I remember seeing a Skyrim mod with an AI NPC that had situational based reactions and responses and I thought that was a genius way to implement it into gaming like imagine the NPCs saying your character's name!
 

PotatoBoy

Member
It's not just an AI generated video of minecraft. Again, from the papers



It's generating a world and understands it. This is no longer the dreaming god AI.

Nvidia's AI expert on the subject







But what does he knows

This AI with spatial awareness and how things should not collide, understanding of physics and how it should react is massive and goes even beyond videos and games, this is something self-driving cars have completely lacked. This will also boost autonomous cars / robots.


You did not watch this video carefully. Scrub it back and forth and watch how the ships move. Especially the one on the right. It looks more like an optical illusion than it does like physics. At some points it almost looks superimposed on the water. Nvidia has a hell of a lot of bridges to sell you if you think this is how a ship behaves.
 

FStubbs

Member
I'm not generally a proponent of AI - I an a "specist" - but there is ZERO doubt this is coming and it will 100% decrease labor costs causing mass layoffs of people which will lower the extremely too large budgets for the AAA gaming space. Of course the "no crunch" FKA "overtime" crowd will protest/strike making it really easy for the suits to decide who is easily expendable.
The suits themselves could get gored. Don't like Call of Duty? A week or two of AI prompting and you have a working clone, just the way you like it.
 

unlurkified

Member
WbOAweP.gif


Edit - also, good luck if you are looking for work and didn’t create the perfectly tailored resume to the position…you won’t even be considered for an interview even if you are the most qualified…AI is a fucking joke and if you support it now, good luck when you legitimately need to talk to a human being in about 10 years for literally anything from customer service to food ordering. It’s sad that we aren’t even standing up against this nonsense right now….you, me, all of us will seriously regret it.
The first almost all AI restaurant just opened in California:



People already aren’t getting enough person-to-person interaction. Things could get straight dystopian.
 

Dacvak

No one shall be brought before our LORD David Bowie without the true and secret knowledge of the Photoshop. For in that time, so shall He appear.
Holy shit, those Sora examples are actually insane.

My prediction has always been that by 2040, most if not all media content will be generated in realtime per user, automatically creating movies/music/games/books based off of your preferences.
 

Buggy Loop

Member
By the way

This means probably by the next decade we'll get an handheld life-like graphics. Gaussian splats and all this AI are much easier to run than what you would typically require if you ran it traditionally.
 

Buggy Loop

Member
You did not watch this video carefully. Scrub it back and forth and watch how the ships move. Especially the one on the right. It looks more like an optical illusion than it does like physics. At some points it almost looks superimposed on the water. Nvidia has a hell of a lot of bridges to sell you if you think this is how a ship behaves.

A year ago mouths were stretching to eat spaghetti, cheeks became meat balls and it had no temporal consistency.

So… another year of this kind of leap and more is a lot of bridges? It’s nothing. It’s a blip.
 

hinch7

Member
Part of the reason I like video games is that I appreciate the effort that has been put into creating art and crafting stories. If that goes away, I am honestly not sure how I am going to feel toward the medium. The same goes for movies. It is a fascinating tech, and I believe it can do a lot of good, but man... ...this is an insane amount of advancement in a short period of time. It will not stop at arts & entertainment. I can see a future where we will have AI and automated bots creating and mass producing material goods.

Family Guy Reaction GIF
If you can't notice a difference between Ai created art and hand crafted one people will accept it. I personally don't care so long as the product(s) and games created are of high quality and if Ai helps games development I'm all for it. Its the same for graphics and real-time rendering for games or otherwise. Eventually the bulk of CG will shift to Ai.
 
Last edited:

The Fuzz damn you!

Gold Member
The insane amount of “I don’t think AI could do this” here in this thread is insane to me. You know what a bunch of other human beings didn’t think could happen outside of a few geniuses over the last century? Windows, Video Game Consoles, TVs, Air Planes,

You know what else a bunch of human beings thought couldn’t happen? Flying cars. Underwater cities. Universal eugenics. Vacuum tube travel. Perpetual motion machines. Reactionless drives. The list is very, very long.

Not all technological end goals are possible, nor feasible, just because they’re aspirational. This tech could maybe make a ‘Dragon’s Lair’ style game on the fly eventually, and it could certainly speed up asset creation, but it’s not going to be able to take a mouse input and give you an FPS with acceptable latency.
 

Moochi

Member
I've been saying for awhile now that this tech changes the future of games. You use low quality assets and model your game, then you apply a specially trained transformer over top. You get flawless graphics of any style you want for practically none of the gpu power costs once gpu design switched to primarily tensor cores and other AI first architectures.
 

mckmas8808

Mckmaster uses MasterCard to buy Slave drives
I'm not sure how much validity this has, but i saw a thread where they were just copying Shutterstock and they showed the Shutterstock comparable.

This still needs a ton of work, I won't believe it until i use it myself. Right now its just stealing images, and putting them together in a puzzle-like way.

edit, this one:


Oh no. So it's not as awesome as it's being presented. I was wondering how this Tech was creating fake humans doing all of those things. But now that I know they are licensing pictures and videos from places like Shutterstock that explains everything.
 

mckmas8808

Mckmaster uses MasterCard to buy Slave drives
I'm not generally a proponent of AI - I an a "specist" - but there is ZERO doubt this is coming and it will 100% decrease labor costs causing mass layoffs of people which will lower the extremely too large budgets for the AAA gaming space. Of course the "no crunch" FKA "overtime" crowd will protest/strike making it really easy for the suits to decide who is easily expendable.

I think the introduction of AI may actually lead to more people being okay with crunch. There is a difference between Crunch and overtime. One is when a person does more work than the Baseline, the other is when a person does more work than the Baseline and either isn't paid for it or is mandatory with very little rights attached.
 
lol months away... haha no

People are reading too much into these tech demos.

production or game development requires a lot of fine tuning so having full controls of how you make these adjustment is very important. And text based input is actually not that intuitive and will cause more headaches for artists. And it also has difficulty recreating the exact same thing multiple times due to the nature of randomly generated seeds.

Furthermore, procedurally generated content is a thing. A lot of games use it to create maps. You don't need AI to do it.

AI will have some uses in the pipeline, such as physics simulations. Something like a fully destructible map that changes based on your actions might become achievable. However, it won't be able to create a game out of thin air. That's not how it works. A video is not the same as actual game you can interact with.
 

kungfuian

Member
The jumps from painting on a flat unmoving canvas to a moving 2d canvas and then to a 3d canvas seems like a natural progression to me. Especially if an AI tool like this is plugged into existing game making software. Feed it enough examples of games, environments, textures, premade assets, photogrammetry shit, then plug it into Unreal and tell it what you want. Easy to imagine a tool like this creating sophisticated 3D worlds (if not full games) infinitely faster than traditional methods.
 
Last edited:
Top Bottom