Is Xbox Cloud Gaming just a data mine for Muse AI?

Rosoboy19

Rosoboy19

Member
This video popped up in my YouTube feed today, I thought he made some good points about Microsoft's possible intentions for the future.

He basically made the argument that, even if Gamepass isn't profitable, it's still valuable to MS' long term AI strategy as cloud gaming can be used to train their AI model Muse about world building and player interaction. I assume the end goal would be for evil AI overlords to replace humans in most game and content creation.

Amazing and scary at the same time.

 
They made kinect to sell your reactions to advertisements . . . they'd 100% do this with gaming.
Everything MS now is just to feed the AI and ML beast.
 
I'm not sure I believe that, but I will say that I'm going to have a hearty laugh when this AI thing doesn't reach the level of utility that M$ and the like are hoping for and the whole bubble implodes.
 
If Microsoft isn't using streamed game video data to train their AI models then they're stupid.

We're gonna see realtime generated video games within the next five years, easily.

I'm not saying I like it, but it's definitely happening.
 
Everyones online activities are being farmed and fed to shitty AI our magnificent AI overlords as we speak, every hour of every day. I would be shocked if MS didn't farm their own data.

I despise it, but I also love the possibilities. From a nerdy d&d rpg standpoint having essentially an in-game reactive dungeon master would be amazing, talk about endless potential. I don't think we'll see anything with the quality needed to allow immersion for a good long while...and in some ways I hope we never do.
 
i saw the video.....I can't wait to see MS/Xbox failing in a catastrophic way with whatever the fuck they want to achieve.


this kind of video reminds me of how the green rats were talking about cloud... cof cod Crackdown 3/ streaming. "Xbox is always ahead, infinite money and massive infrastructure." ... yet. they are not really where they are supposed to be (according to their size)
JUaLI8OdX9RPbIHR.jpeg


in relation to this "Muse" thing ... i mean, don't you think everyone is going implement AI in the workflows one way or another... like Epic doing their own thing within Unreal Engine? (AI. chat NPCs to begin with)

Sony, Nvidia, (every other publisher ) having their own shit too?

the only thing I find intriguing is going to be the divide between "talentless freaks" using AI to make scam games and talented studios implementing AI to make bigger, better, more sophisticated and original games efficiently and faster. (in the console space)
 
ChorizoPicozo said:



even video generation still is incredibly limited and inconsistent, now imagine videogames
Click to expand...

I agree, but even 9 months ago is almost an eternity in AI progress. I just feel like the tech is going to progress faster than most of us realize, so I'm mentally preparing myself for that.

If it doesn't, that's awesome. But it really seems like it will.
 
Rosoboy19 said:
He basically made the argument that, even if Gamepass isn't profitable, it's still valuable to MS' long term AI strategy as cloud gaming can be used to train their AI model Muse about world building and player interaction. I assume the end goal would be for evil AI overlords to replace humans in most game and content creation.
Click to expand...
Watching how other play helps you to play game, not create one. It's not even help evaluate game as its impossible to determine just watching people playing.
And what's the point in replacing players?

Dacvak said:
If Microsoft isn't using streamed game video data to train their AI models then they're stupid.
Click to expand...
The problem is that this data is pretty useless as this training ain't sellable

Dacvak said:
We're gonna see realtime generated video games within the next five years, easily.
I'm not saying I like it, but it's definitely happening.
Click to expand...
No chance
 
Last edited:
All these companies (gaming etc.) seems to have become a lot more predatory as of late. Not saying all of them never have been. But it seems like from about 2021ish companies have really become predatory to the extreme.
 
I have no doubt this is a motivation for Microsoft at large, and Phil is doing what he can to "align the Xbox brand with the broader strategic goals of the company" - but that doesn't mean shit if the AI that gets developed sucks and doesn't provide enough value to developers to become a profitable revenue stream. I just don't think it will get there. In fact, I would assume Phil already knows that, and he's just playing along to keep his paycheck coming.
 
Felessan said:
And what's the point in replacing players?
Click to expand...
I can imagine at least 2 functions for AI use

play testing. An AI that tests your game 24/7 before release at every stage of development trying every path with every possible configuration should find even more bugs than badly payed, bored humans having a hard time playing games in a trash alpha condition.

struggling mp modes to fill lobbys with AI players that actually offer human like skills, for better and worse.
 
If there isnt a part of the EULA for every online game that allows them to record user inputs I would be shocked.
 
crackajack said:
I can imagine at least 2 functions for AI use

play testing. An AI that tests your game 24/7 before release at every stage of development trying every path with every possible configuration should find even more bugs than badly payed, bored humans having a hard time playing games in a trash alpha condition.

struggling mp modes to fill lobbys with AI players that actually offer human like skills, for better and worse.
Click to expand...
What does both have to watching zillions of hours of others playing irrelevant games?
AlphaGo/DeepBlue didn't watch millions of video people playing cards or board games, they played their respective game (Go/Chess) for thousands of matches and get good in them.

PS. Playing games is even completely different AI model that learning by playing and not watching. Models that watching others are to imitate others and in gaming it's not relevant (or can be done with significantly less resources as UT custom bots showed 20 years ago)
 
Felessan said:
What does both have to watching zillions of hours of others playing irrelevant games?
AlphaGo/DeepBlue didn't watch millions of video people playing cards or board games, they played their respective game (Go/Chess) for thousands of matches and get good in them.

PS. Playing games is even completely different AI model that learning by playing and not watching. Models that watching others are to imitate others and in gaming it's not relevant (or can be done with significantly less resources as UT custom bots showed 20 years ago)
Click to expand...
of course you can train AI indepentantly, but I assume AI would get desired results faster if it is trained on how humans play. AI is supposed to emulate players and not emulate something else in my use ideas.
DeepBlue did analyse actual games... the ruleset was set by humans though and not learned by training and playing as AI chess engines do now.

what was called AI in UT was certainly not AI in the current sense.
 
Last edited:
crackajack said:
of course you can train AI indepentantly, but I assume AI would get desired results faster if it is trained on how humans play. AI is supposed to emulate players and not emulate something else in my use ideas.
Click to expand...
It pointless to emulate human players.
For testing you want a perfect automation that can "go for every route", not an average Joe running same route 95% of the time.
And for bots it's pointless to make them really good/lifelike, for psychological reasons it always will be more fun to play against living persons and bots always would be sub-par choice so they need to be just good enough. So no one really put any efforts into improving them for long time already, be it AI or not, as it doesn't bring value to the game.

And remember, training set will consists of random players sample and most players are casual with corresponding level of skills. And as there is no feedback loop, unlike actual playing game, AI learning from videos would imitate the most common scenario and it's not what people would like to see as a result of training.

crackajack said:
DeepBlue did analyse actual games... the ruleset was set by humans though and not learned by training and playing as AI chess engines do now.
Click to expand...
It's to speed up training but it's a lot of extra efforts (and thus expensive), so this part now often omitted unless you want to teach AI something extra rare that would take ages for him to learn by itself.
Not an usecase for general gaming applications
 
And that's why Google Voice was free to use for years. They used it to train speech recognition and generation. Or maybe it was just called Google Phone. I don't remember. But it let you make calls basically for free. Turns out, we were the product, just as the adage goes.
 
Last edited:
You must log in or register to reply here.

Similar threads

Xbox unveils Muse, an AI for gameplay ideas.
2 3
119
7K
ChorizoPicozo replied
XBox Producer Suggests Recent Layoffs Use AI For Emotional Support for Losing Jobs to AI
2 3
136
1K
Landr300 replied
LTTP: Xbox Cloud Gaming (Via Samsung TV)
30
2K
WX3 replied
Xbox's Cloud Gaming Expanded to Game Pass Core & Standard Subscribers (Previously Ultimate only)
30
133
Quasicat replied
Xbox cloud games will soon follow you across Xbox, PC, and Windows handhelds
Trailer Platform 
15
622
Hashassin replied
Top Bottom