Xbox pushes ahead with new generative AI | Hacker News Reader

Xbox pushes ahead with new generative AI

wired.com

28 points01-_-33 comments

From the Xbox press release (https://news.xbox.com/en-us/2025/02/19/muse-ai-xbox-empoweri...):

> Today, countless classic games tied to aging hardware are no longer playable by most people. Thanks to this breakthrough, we are exploring the potential for Muse to take older back catalog games from our studios and optimize them for any device. We believe this could radically change how we preserve and experience classic games in the future and make them accessible to more players.

> To imagine that beloved games lost to time and hardware advancement could one day be played on any screen with Xbox is an exciting possibility for us.

To me this shows that they're just bullshitting to pump the stock price. There's no way that using generative AI to port nontrivial games based on gameplay videos and recorded inputs would produce anything more than maybe a tech demo at best. You wouldn't get anywhere near the accuracy of emulation or a direct source port, but saying "we're hiring emulation experts to improve our backwards compatibility" doesn't juice the stock like "we're using generative AI to preserve games" does.

Exactly. Being able to play old games is a solved problem. They might as well say that they need quantum computers on the blockchain to help people play pac-man.

Tl;Dr a hammer looking for a nail.

Don't speak on my behalf. I'm a gamedev and I want this. Claiming faster prototyping is wrong because it's "as much about the journey as the result" is just ridiculous. I want to quickly test mechanics and couldn't care less about the intrinsic motivation.

cube(getDimByType(type), getRandomColour) worked fine for a million years and still works in Minecraft.

The last thing needed for a game jam are tripple aish models mixed with pixelart?

The answer to your question is no.

Well that should depend entirely on whether or not it actually works, yes? Either it makes it cheaper and/or faster to develop new games, or it doesn't.

I was kinda hoping for better games, not cheaper or faster.

With powerful tools you get both easy slop and cheaper masterpieces.

Authorart without the (Its500$LetsHaveAMeetingAboutThat)Gremium of meddle management art critics meeting. Scandalous!

Nobody is sculpting better than Michellangelo with angle grinders though.

Yeah but the analogy goes, if Michelangelo could snap his fingers to remove the first 80% of material on a chunk of marble, he could focus on the details and get more done, or have more time to perfect.

Generative AI is best at things it has been trained on the most, so the most repetitive, rote, boilerplate, tasks become much quicker.

> Yeah but the analogy goes, if Michelangelo could snap his fingers to remove the first 80% of material on a chunk of marble, he could focus on the details and get more done, or have more time to perfect.

No. If that was a valuable thing that could be delegated, he could have "snapped his fingers" and have assistants do it. Or maybe sculpting isn't all about the later detail work like you assume, and removing that first 80% also requires skill and can't be delegated.

'Better'? No, no, that's not AI bubble thinking. "Crap, but very large volumes of it" is what is in right now.

> Well that should depend entirely on whether or not it actually works, yes? Either it makes it cheaper and/or faster to develop new games, or it doesn't.

What if it makes them cheaper, faster, and worse?

Then I suppose someone would miss their sales targets.

Is it so bad that nobody buys, or does the cost of fixing exceed the initial otherwise cost?

I have no idea what's going on at Xbox. I'm not much a gamer anymore, but like to read about it still.

The latest versions are a sales flop. The mid cycle refresh was canceled even(or whatever Brooklin was). And agreeing to release games on PS seemed an acknowledgement of that.

A lot of folks thought this indicates that MS would get out of hardware here, but this seems to be investment in it.

Are these just unsubstantiated rumors, or is this the right hand not knowing what the left is doing?

Microsoft has always been at war with Microsoft.

I think what they want to do is create the best Xbox console to run their games on, while also making money from their games across all platforms. Its not necessarily dumb they just have to handle it well (which they probably wont)

As someone that has played PC games since 1990, the problem is XBox has turned into a cross platform publisher after all those studio acquisitions, especially ABK.

So big mamma Microsoft, doesn't want to see only a fraction of sales on XBox hardware and PC, which given Microsoft's history always had a place alongside XBox, when they can go big (in money terms), across all platforms.

So XBox business unit has a big problem now, trying to assert the console still matters, when Microsoft as a whole, wants the big bucks more than anything, and doesn't really care about the console, rather XBox as a platform, Netflix for games.

> A lot of folks thought this indicates that MS would get out of hardware here, but this seems to be investment in it.

I mean, it's a not-to-be-missed opportunity to make a press release containing the word 'AI'. They'd be fools to turn that down.

Relevant recent discussion on "Minecraft with object impermanence", which is an experiment along the same lines as this: https://news.ycombinator.com/item?id=42762426

Worth it for this alone

>Among the various use cases for Muse that Microsoft outlines in its announcement, perhaps the most intriguing involves game preservation. The company says Muse AI can study games from its vast back catalog of classic titles and optimize them for modern hardware.

"Optimize" meaning super shitty generative upscaling from VGA to 4K or something? Yeah, I’ll pass.

Oh, so that's where the bad AI generated content for AOM came from!

Can see the potential uses for prototyping, like if you fed in some concept art as the initial frame, but struggle to see the applications for game preservation. I feel the result would be more demanding to run than traditional emulation, slightly off in how it plays, and missing any content/interactions that wasn't covered in the training videos of the original game.

There are very valid use cases that don’t involve firing artists. Plenty of upscaling / remasters benefit from this tech.

If we think further, applying AI to something like RTX Remix would allow a lot of progress in keeping older games vibrant.

There are downsides and poor examples for sure, but if done well this could also help.

The PS5 is dominating the market for good reason even with partial AAA title coverage.

The Xbox platform answer was to release Clippy v2a =3

https://www.youtube.com/watch?v=5rAOyh7YmEc

True regarding hardware console sales, and playground arguments.

However the hard truth many XBox console owners are having problems to accept is that XBox Platform end move is to become Netflix for games via Game pass, and being a 3rd party publisher everywhere.

How many first party titles where at last Playstation event versus 3rd party?

The industry is not in a growth-cycle right now, and many studios preclusion to squeeze people is failing to rally consumer goodwill. It is a retail condition that initially opened the markets for Microsoft in the first place.

Some platforms are indeed already having a de-listing problem with several popular titles. The value proposition of disposable hardware has certainly changed product lines. =3

Console exclusives have long been a complaint from gamers. Now Microsoft has essentially eliminated those exclusives and people are upset? It's bizarre.

"nobody will want this"

Reply on news.ycombinator.com