Microsoft’s Copilot Gaming Experience is haunted and pointless

Tech companies are really pushing generative AI for some reason. The technology has a lot of practical uses, but companies would rather wave their dicks around to show it off. Their dicks, in this case, being braindead chatbots and dangling carrots in the form of generated art. The latest in these flashes of useless genitalia is Microsoft’s “Copilot Gaming Experience.”

The idea of being able to spontaneously generate a video game based on a prompt is something that has companies salivating for some reason. I want to point something out: if something takes no effort, it has no value, so I’m not sure what they’re excited for. If anyone can generate a game, no one will be paying for them, so…

Anyway, that’s not entirely what Microsoft’s Copilot Gaming Experience is. Instead, it’s Quake 2. Now, you might be saying, “Quake 2 already exists; I own it fifteen times across multiple platforms.” I know. Me too. But it’s not so much that this allows you to play Quake 2, it’s that you get to play Quake 2 in a way that sucks up already depleting resources and spits them out in the form of a nightmarish electric sheep. That is to say: what androids fever dream of.

A TOASTER WITH CHRONIC TRAUMATIC ENCEPHALOPATHY

I don’t know where to begin with this. On paper, as I’ve said, it’s the first level of Quake 2, and the toaster running this should have been given that paper, because it often forgets. If you are really diligent, and already familiar, and also very lucky, you might be able to navigate to the first loading screen where the experience ends – after all, it’s only down a couple corridors. However, you probably won’t have fun, and you’ll most likely fall victim to the computer’s forgetfulness.

It’s easy to experience; just look at the floor. When you look back up, you’ll probably be in a different spot in the level. Try swimming. It will make the experience completely wig out, because it can’t tell the difference between one area with water and another. At one point, I found myself inescapably teleported to a hidden part of the level that I couldn’t escape. Copilot is, ironically, completely inept at navigation.

There are enemies, but they show up as unsettling ghostly blurs. Sometimes, you might even find yourself in a moment where you’re seemingly chasing their phantom image down a corridor. They can attack you, but the AI struggles to keep track of your health and armour. You can kill them, but don’t expect Copilot to remember. It can’t even remember the room you’re in.

“Concretely, we introduce a state-of-the-art generative model, the World and Human Action Model (WHAM), and show that it can generate consistent and diverse gameplay sequences and persist user modifications—three capabilities that we identify as being critical for this alignment,” says Microsoft despite their model demonstrating that it’s completely incapable of these things.

ONE BLEAK FUTURE

I’m not sure who this is supposed to impress. I’m guessing it’s everyone who already thinks this technology is amazing. That’s true, I supposed. Generative AI is amazing technology, but it’s wielded primarily by feckless jackanapes who can’t see the wall in front of them. The same sort of feckless jackanapes who are deluded enough to think that this is an impressive demonstration of the future they envision.

But feeding a near 30-year-old game into a computer and having it shit out a confusing mirage is pointless. It wouldn’t exist without the development expertise of the humans who made the game, and it only makes it significantly worse. It adds nothing, and takes us nowhere.

This is frame generation without the frames. The big new graphics technologies (DLSS and FSR) have generative AI “guess” the image that would occur between locally generated frames to make the game appear smoother. Smoother as in a higher FPS, not in terms of texture, because the noise added in each image takes a lot away from the overall picture, but the idea is sound. Frame generation has promise, even if I think the effort outweighs the advantages. The value of frame generation is at least, I feel, a worthwhile debate. This Copilot Gaming Experience, however? I don’t see how this is valuable.

INHUMAN MACHINES

Generative AI has already made its way into games. In some cases, I can at least see where there’s value in current applications, as it allows some smaller developers to create quick assets to fill in gaps. Most often, I see it as posters on walls. I don’t agree with the usage, but I see how it might make sense.

There are other games that have been more-or-less completely assembled using AI, and they invariable suck. I am open to the idea that maybe advances in the technology will make them not suck, but I remain doubtful. But then, that still doesn’t really explain where the value comes from. If something doesn’t take any effort, there’s no value, and you can’t sell it.

At best, if I really strain my imagination, you’d be able to use something like this to take a pre-rendered game and turn it into something playable. Like, if you fed WHAM Myst, it might be able to turn it into a game where you can walk around and explore. Never mind that a full-3D Myst has already been done. If you fed it Google Street View, it might be able to guess what’s in people’s backyard if that’s helpful. Just spitballing here. Giving the benefit of the doubt that there will one day be a practical use of this hog swill.

“CONSISTENT AND DIVERSE”

I guess, maybe, the end goal is that you don’t need other people’s games? You pay a fee to an AI company and you get infinite bespoke gaming experiences. Any game you want, a toaster can create it for you. That’s weird. That’s like saying that you don’t need sex because you can just masturbate. Infinite fantasy and the same payoff. Or maybe that you don’t need food because multivitamins and wallpaper paste give you all the nutrition you need.

I wouldn’t have asked a toaster for Chulip, because it’s not something I would think of. My favourite games aren’t ones that I envisioned. I didn’t think that one day I’d want to play a game about rampantly promiscuous kissing; someone else drew that out of themselves, it became distorted as they pushed the concept through various obstacles to produce something that worked, was fun, and reflected a part of their imagination, and then they offered it, and it has lived in my brain since.

“Extrapolating from our use case focusing on a single 3D video game, we can also get a first sense of how powerful future models will be in allowing teams of human creators to craft complex new experiences,” says Microsoft. Fucking what? You regurgitated Quake 2, which was made by “human creators.” You didn’t “craft [a] complex new experience.” A computer can’t do that because it will only be able to give what we want to have, and not what people have to give. This doesn’t give a sense of anything except how much time and money you’ve wasted.

BEYOND WHAT HUMAN HANDS COULD BREAK

It’s all conjecture right now, anyway. In the present, we have the Copilot Gaming Experience, and it is fucking sad. It’s a toaster’s brain on drugs; a nightmarish distortion of something familiar. A rag drenched in the genetic fluids of shortsighted investors.

If there’s any value in the product as it is now, it is, unintentionally, a unique experience. It’s a specific sort of horror; one that is both uncomfortable and inhuman. The horribly stiff movement and significant amount of lag. The complete lack of any sort of tactility or place. This is a whole new level of terrible. It’s broken beyond what human hands could possibly break.

If anything, it’s something that a human could take as inspiration and turn into a new sort of atmosphere. Except the whole, “look one way and everything behind you changes” thing has already been made. So has the blurry enemies, for that matter. Fucking toasters can’t even accidentally do anything new.