Microsoft recently announced Muse, "a generative AI model of a video game that can generate game visuals, controller actions, or both". And somehow, Phil Spencer of Microsoft Gaming thinks it will help game preservation.
Speaking in a video that has been set as unlisted (hidden) on YouTube, but included in their Xbox news post, Phil Spencer of Microsoft Gaming goes off talking about how this somehow might magically help game preservation. No, I don't know how it would either, that's not how it works. Getting a generative AI to hallucinate games is not really preserving anything.
Anyway here's what Spencer said: "I'll tell you one of the things that I get excited about. You know, one of the things we care a lot about at Xbox is game preservation. And I think about an opportunity to have models learn about older games, games that were maybe tied to unique pieces of hardware where that engine on that hardware, is kind of time will erode the amount of hardware that's out there that can actually play a game. But you can imagine a world where from gameplay data and video, that a model could learn old games and really make them portable to any platform where these models could run."
Continuing Spencer said: "We've talked about game preservation as an activity for us. These models and their ability to learn completely how a game plays, without the necessity of the original engine running on the original hardware opens up a ton of opportunity."
The video can be viewed below:
Direct Link
At some point, this whole "AI everything" thing will come crashing down. But until then, we will see more and more executives dreaming up things that won't happen. Game preservation is absolutely vital, and there's so many ways to do it correctly, but getting an AI to dream one up that it thinks is accurate is definitely not it.
You can also read more about Muse in the Microsoft news post.
Article taken from GamingOnLinux.com.
pleasereadthemanual a day ago
Mohandevir a day ago
I trust in you all to correct me, if I'm wrong, which is highly probable.
I'm not surprised that M$ is trying AI even in the XBox department. From the short term perspective it would be bad for them to not do it.
I can’t imagine the amount of trial and error it would take for an AI to figure out such a complex problem, though. Perhaps it’s more likely that it could be used by human reverse engineers to probe and test obscure functions faster.
Tadas-Estidal a day ago
not to be a doomer, but this will only get worse.
good thing is - we have enough indie and other decent studios to make us great games.
I have absolutely no clue about AI, but I'm wondering if it couldn't be used, for old games that we have access to the code. Kind of feed that code to an AI that will make it compatible with modern hardware?
That's a far more feasible scenario than the absolute word soup Phil is spilling above. The challenge would be training the AI on appropriate hardware data and interfaces so that it can transition the code from one architecture to another. Still a far cry, in my mind, but at least you can see the bare bones of how you might go about it.
Showing an AI videos of the game and getting it to recreate the game... and then calling THAT preservation? Laughable. Cringeworthy. Deluded.
R Daneel Olivaw a day ago
At some point, this whole "AI everything" thing will come crashing down. But until then, we will see more and more executives dreaming up things that won't happen. Game preservation is absolutely vital, and there's so many ways to do it correctly, but getting an AI to dream one up that it thinks is accurate is definitely not it.
You put it perfectly.
Also. Can the crash please happen, like, tomorrow? The whole deepseek thing is helping, but I'd love to just turbocharge the collapse so we can be through and done with all the absurdity.
Also. Can the crash please happen, like, tomorrow? The whole deepseek thing is helping, but I'd love to just turbocharge the collapse so we can be through and done with all the absurdity.This entire cycle just reminds me so much of RPA from around the 2010's. Same bullshit optimism from execs about "staff reduction" (or, if trying to be politically correct, "staff efficiency") and same underwhelming, erratic, difficult-to-reproduce, difficult-to-troubleshoot results.
The worst thing about RPA was that the slightest change to your processes would break it (even when those changes were outwith your control, such as when accessing data via a website, or API). I suspect the same will be true if you introduce AI.
Anyway, there's STILL companies banging on about RPA today, so don't get your hopes up on the AI crash being all that dramatic. Companies have already poured too many billions of dollars into it for it to just go away, sadly. Gartner are predicting hundreds of billions of spend this year and next. The crash will come, I'm sure, but it'll more likely along the lines of the well known "Gartner Hype Cycle". The crash will only realign expectation to reality, then we'll actually see AI being used where it's actually effective, and we'll stop seeing utter crap like Phil was spraying, above.
Generative AI could become another alternative under our current efforts of game preservation, but never the sole way to achieve it. Otherwise, it's not even emulation anymore, and prone to errors and delusions like our current models are.
It looks much more promissing as an iteration tool during game development instead of an attempt at game preservation, as it gives you examples and fast feedback of an idea you might have for your current build, when used responsibly in addition to other tools and dev paradigms leading everything.
It sure would be very interesting if someone ended up making a machine-learning tool that can mass reverse engineer software.
As someone who has reverse engineered old games... yeah, that's not going to happen.
Perhaps it’s more likely that it could be used by human reverse engineers to probe and test obscure functions faster.
We don't need AI for that.
IDA can already identify tons of standard library functions by pattern matching against a library of known functions. With that, you already strike off like 90% of all simple functions. The rest you can figure by yourself pretty quickly.
The issue is the non-simple functions that depend on a large context, of passed in parameters or, even worse, global state. There, being able to run the game and attaching a debugger to catch the state helps. AI doesn't do anything there.
Maybe you'll now say, well, let's attach AI to the running the game and have it analyze it like that... well, that just boils down to the halting problem, so we know that doesn't work.
So...
You know what would actually help game preservation? Companies being mandated to 1) keep their source code around and 2) openly publish said source code with a FLOSS license. And then also doing that for the source history, asset sources, documentation, internal process documents, etc.
Patreon
PayPal
How to set, change and reset your SteamOS / Steam Deck desktop sudo password
How to set up Decky Loader on Steam Deck / SteamOS for easy plugins
See more from me