Google's AI Just Built a Game in Seconds… But Is It Any Good?

[@kogekisaru]

Genie 2 isn't making games. It is an AI generated video that dynamically accepts keyboard input as additional prompts while the video is playing. You can't prototype anything here, because there is nothing to spit out.

[@martinjakab]

Prototyping in this case means validating ideas. If this ever becomes good, it will be more than capable to that.

[@zakjj2233]

Definitions wont change the reality

[@blurchill]

That’s the same thing as a videogame. At the end of the day it’s just frames drawn on a screen that reacts to inputs. Genie 2 is basically cloud gaming. It’s not running locally but actions are happening based on your inputs. Sure you could say “hey, what about saves, the items in the game, achievements?!” But those are just extra stuff. At the end of the day, yes, it’s a video game. And if I put you in a room with something made by “Genie 10”, I bet you wouldn’t know the difference. Edit: I mean it’s literally in the name of it. VIDEOgame. It’s technically a video that is interactive (running at X fps).

[@Selrisitai]

Remember that this "A.I." is still just a BASIC model that uses averages of what it's been trained on. "If I see this, then this is what usually would be next." I don't think this concept can ever go particularly far. A whole new technology will need to be invented to make this work.

[@AlleyKatPr0]

You got UE5? Open a new project, turn on python scripting, and then ask your favourite AI to generate UE5 assets using the 'asset factory' for, like, a BP or a 'Master Material' all complete with nodes in the node graph. Tutorials are now, worthless.

[@DailyShit.]

Parts of it can be included into bigger engines though. But always need to watch out for performance.

[@Selrisitai]

@@AlleyKatPr0 But these are standardized things, right? That's why it can technically write an entire book, because it will suck, and I don't think an improvement on the technology it has now will make it stop sucking because it doesn't seem to fundamentally work in a way that would allow it to stop sucking at creative thinking. In fact, what you just described is basically just a text command for a standardized process, as far as I can discern.

[@DangerDurians]

“We’ve got a brand new idea that’s never been seen before that’s really specific, make us a prototype” Genie 2 : “sorry, can’t replicate a mechanic that I have no reference for” spits out generic 3D third person walking sim, but the world is cool (and blurry)

[@mgodoi3891]

We, human don't create nothing either. We just grab everything throw it in a mixer and come up with something "new". I'm a kind of an artist and soon i will be replaced too. It's inevitable

[@professorfroopynoopers1171]

It's always to replace, always. The C-suite for any company will always always look to make investors happy first. Investors care about one thing, money. Not artists, not consumers. To humor any other possibility is an insult to human intelligence. We will be replaced, just a matter of time. Unless people stand up for themselves and stop treating these developments like they're not being intentionally designed to replace humans.

[@salmadys]

I already have more great games a year than I have time to play them and these were made by real creative humans. All these AI prototypes seem like empty promises that barely reach the current state of the art gaming for mere 12 seconds, why waste my life in all this slop? And instead of making tools to help real artists to automate the boring parts of game development; Retopology, rigging, animation, rendering in 2d art or bridging the gap between 2d and 3d with automatic depth maps or normal maps. The industry is just pouring money in useless magical solutions in an attempt to attract more investors into their little scam. In the end these AI solutions are as uncreative as the tech bros pushing for them.

[@BryanMaynardsGoogle]

What I see of Genie 2 in this video looks like nothing more than smoke & mirrors; real geometry doesn’t do what it looks like that stuff is doing. This looks like clever use of video to me. If/when this can interoperate with tools like Unreal, Unity, Godot, etc I’ll give it some attention. Until then, I’m filing this under “Lies trying to keep [insert corporation here] relevant.”

[@TheJofurr]

The biggest problem with AI art is that I've never seen it make anything really interesting on its own, the output is incredibly generic and bland. As a tool in the hands of an artist it can improve efficiency but it still needs human direction and creativity, without that it's just an aggregator that's going to expedite the downward spiral of 'design by committee' that the corporate profit motive already puts everything on.

[@alfistibrasiliani]

well, it's the same with tradicional artists and illustrators! No one is born ready to draw. Everybody needs to get references and influences. 99% of the artists follow a know style or aesthetic. Only a handful have original work.

[@Kio_Kurashi]

I could understand using AI to create portions of a world. Example being a single donut model or a basic framework for a level. But it shouldn't be used to try to make everything at once. And this is also assuming the AI was advanced enough to not be a crappy buggy mess.

[@hollowedboi5937]

I'm just tired of crap happening in the world as a whole, including AI. As an artist, I think the text based stuff helps to assist in decision making and research, but that's kinda the extent. Generative stuff nah mate I'm good. I wanna create not generate.

[@pismodude2]

It "helps" but ultimately many of us stopped using it after it made impossible suggestions with confidence. When confronted with the reality that something it suggested wasn't possible technically it just immediately goes "Yes you're right I was wrong" but then continues to make the same suggestion. It's like having a yes man who gives answers you want with confidence and no regard for whether it's sinking your ship because it's not thinking it's just responding using public data framed in an unsourced and confident statement. Not a huge fan of it after running into this issue so many times.

[@jj1746]

doesnt look like its useing any real assets, just kind of an ugly movie u can move around

[@pismodude2]

It seems like it could be combined with a visually simple game and taught to map the good-looking things to certain objects. Kind of like DLSS or frame generation except it's turning a red square that jumps up and down into a person. Depressing to imagine coding an entire game with simple stand-in objects and letting the executives and marketing department give commands to an AI that handles the art direction for your project though...

[@doudar41]

In the "creative" line of job general client requirements are "I want a professionally looking/sounding product that looks like "X", without (c) problem". If you want to be this "X" you need to sell your creation first to get these clients attention. So they will be sure that creations which resemble yours can be sold. Even without AI this problem was obvious. AI just makes this problem even more obvious. This is the highest level of making an impression instead of creating something real. Last years of my job as a sound producer for radio and TV in cases when people wanted "something like that" I told them that I could teach them how to do this or advise a person to work with instead of doing a stupid job. AI doing exactly that, doing stupid jobs for a small price. There are a lot of people who are after this kind of thing. Only thing that could change this, if clients start to lose money by making AI content and start to talk to creators, look for real creators, put money into searching talented people who don't have a tick tok account. But I don't think this is our nearest future.

[@zakjj2233]

After knowing ai exist, learning shit is irrelevant

[@forestpump3]

This feels like a fireship video

[@Micha-Hil]

"Genie 2 can make game development super easy!" Okay make me an online multiplayer game then.

[@MrSkullhunterxx]

am i crazy or is this ai text to speech ...

[@Zoxfr]

I guess I’ll Mark my spot

[@Soguwe]

Making a world is not the same as making a game.

[@Badguy292]

DAMN, you're just gonna AI Generate me to subscribe? Good thing I'm already subscribed though.

[@justice3007]

A tool doesn't replace expertise

[@felman87]

Companies would be wise to remember the video game crash of 83 and how that happened because so many companies flooded the market with crappy titles that no one wanted to buy games anymore.

[@makotoseto6376]

This would be groundbreaking if it has the potential to optimize games and fix glitches, idk if it would look good graphics wise though

[@thentheric6361]

I think this will be great for prototyping, especially enacting a GDD's "First five minutes of gameplay" section, which can make or break an idea. It can help rapidly iterate on the "Would this be fun?" trial-and-error part. We have to understand that AI is still a productivity tool, like photography for capturing reality, or 3D printing for assembling engineering prototypes. These automated solutions will always produce cheap solutions rapidly, but they will lack the detail, refinement and unique intent and purpose of a human-provided, or human-refined product.

[@myxsys]

What if you have a totally unique game mechanic idea? How will AI generate a feature it has never been trained on?

[@thentheric6361]

@myxsys Then it's a hit or miss, and you've only wasted 3-4 description refinements on it, probably coming to understand what you need to do from your end. It's a good indicator that your idea is very novel, and it will take a lot more work to properly design. AI is still a tool for a purpose, and it will stop being useful when we go beyond the patterns it can reliably work with, and it will fail you if you don't know how to engineer your prompts correctly. You need to do a budget analysis to even see if buying/downloading an AI and a model, and setting/retraining it is what you need, or if you're better off with a paid service. Factories did not nullify our work, they just made it more resource-efficient at the sacrifice of end quality. A suit just off the shelf is never as good as one you had fitted afterwards, or tailored onto you from scratch. (Of course, AI can also simplify the measurement and design, if trained for it, but again, you need the right set of skills to use it.) Returning to the original topic, game platforms will not allow AI slop to flood their shelves, because people will simply go and look elsewhere for quality, playable games. Or if not, that's pretty sad, and shows that you're probably working for a niche audience, because potential players are tasteless pigs who only want the next reskin of "match three" or "giantess-vore-fetishist number-comparison quest" brain rot.

[@rothko459]

I think my main complaint with this type of gen AI is that there's no way to control anything about it. You can indeed type 'make a puzzle game with cyberpunk aesthetics' and just get a mess that would be so useless you might as well do it from scratch. Like once you get past the 'wow it made something from a text prompt' novely it just falls flat.

[@richardjecooke8036]

This video is not true as far as I understand Genie. It doesn't create "3d worlds", it creates an image based on a previous image, and loads those in succession to create video.

[@TylerDane]

It's a really bad look to be giving this kind of slop any attention whatsoever

[@d-pod_L]

If they perfect this tech, why would anyone even buy games, when you can just make exactly what you wanna play yourself?

[@definitelynotnick2454]

Yeah like how Dreams put the game industry out of business /s

[@xXESproductionsXx]

We sbould not be striving for art to br randomly generated

[@A120AMIR12]

It feels far from polished But it's like the start of game dev history it self Back then it was falling apart until Nintendo and other company saved it

[@zakjj2233]

Wait for open ai game maker

[@AlleyKatPr0]

It's very simple - those who play good games will be the ones who can communicate with AI, so as to get the AI to generate the game we want to play. The more accurate the description the gamer gives the AI, the more closely the AI will give the player the game they want. No one will be giving money to games developers, they will just have a subscription to AI...BUT THEN, AI will become soooo good, it will be duplicative by anyone, so even the subscription model falls apart. Did you really need a video to explain this?

[@pawan4920]

🗿

[@Alice.Kreuzfeuer]

Nowadays Steam is full of copy-paste Unreal Engine asset flips. Tomorrow it will be full of this AI-generated bullshit... Not a big deal 🤷‍♀

[@r99716]

won't be anything more than a toy for awhile... but if you see this and aren't amazed then your inner child is dead