https://arstechnica.com/information-technology/2024/08/new-ai-model-can-hallucinate-a-game-of-1993s-doom-in-real-time/
Dobos: "Why write rules for software by hand when AI can just think every pixel for you?"
"The potential here is absurd," wrote app developer Nick Dobos in reaction to the news. "Why write complex rules for software by hand when the AI can just think every pixel for you?"
"Can it run Doom?"
"Sure, do you have a spare datacenter or two full of GPUs, and perhaps a nuclear powerplant for a PSU?"
What the fuck are these people smoking. Apparently it can manage 20 fps on one "TPU" but to get there it was trained on shitload of footage of Doom. So just play Doom?!
The researchers speculate that with the technique, new video games might be created "via textual descriptions or examples images" rather than programming, and people may be able to convert a set of still images into a new playable level or character for an existing game based solely on examples rather than relying on coding skill.
It keeps coming back to this, the assumption that these models, if you just feed them enough stuff will somehow become able to "create" something completely new, as if they don't fall apart the second you ask for something that wasn't somewhere in the training data. Not to mention that this type of "gaming engine" will never be as efficient as an actual one.
Apparently it can manage 20 fps on one "TPU" but to get there it was trained on shitload of footage of Doom. So just play Doom?!
Shhhh! Are you nuts? People are going to start realizing this is another tech bubble, like Blockchain...
/s
To be fair, half of the AAA gaming industry is all about trying to clone the latest successful game with a new coat of paint. Maybe using AI to make these clones will mean that the talented people behind the scenes are free to explore other ideas instead.
Of course in reality, it just means that the largest publishers will lay off a whole lot of people and keep churning out these uninspired games in the name of corporate profits, but it's nice to dream sometimes.
I mean, you've never seen a purple elephant with a tennis racket. None of that exists in the data set since elephants are neither purple nor tennis players. Exposure to all the individual elements allows for generation of concepts outside the existing data, even though they don't exit in reality or in the data set.
Ok.
Try to get an image generator to create an image of a tennis racket, with all racket-like objects or relevant sport data removed from the training data.
Explain the concept to it with words alone, accurately enough to get something that looks exactly like the real thing. Maybe you can give it pictures, but one won't really be enough, you'll basically have to give it that chunk of training data you removed.
That's the problem you'll run into the second you want to realize a new game genre.
There are more forms of guidance than just raw words. Just off the top of my head, there's inpainting, outpainting, controlnets, prompt editing, and embeddings. The researchers who pulled this off definitely didn't do it with text prompts.
Obviously.
But at what point does that guidance just become the dataset you removed from the training data?
To get it to run Doom, they used Doom.
To realize a new genre, you'll "just" have to make that game the old fashion way, first.
But at what point does that guidance just become the dataset you removed from the training data?
The whole point is that it didn't know the concepts beforehand, and no it doesn't become the dataset. Observations made of the training data are added to the model's weights after training, the dataset is never relevant again as the model's weights are locked in.
To get it to run Doom, they used Doom.
To realize a new genre, you’ll “just” have to make that game the old fashion way, first.
Or you could train a more general model. These things happen in steps, research is a process.
You are completely missing what I'm saying.
I know the input doesn't alter the model, that's not what I mean.
And "general" models are only "general" in the sense that they are massively bloated and still crap at dealing with shit that they weren't trained on.
And no, "comprehending" new concepts by palette swapping something and smashing two existing things together isn't the kind of creativity I'm saying these systems are incapable of.
What kind of creativity are you talking about then? I've also never heard of a bloated model. Which models are bloated?
It's basically like. Someone drawing a picture. Then watching the buttons you're pressing on a controller. And then drawing a new picture. And based on the game that they think you're playing in their head trying to guess what the next picture ought to look like. With no error correction and no conceptualization other than what the next picture should look like.
The... many limitations of this is the inability of image generators to rationalize 3 dimensional space. It can only approximate it based on what it thinks should appear on the screen. It lacks any ability to keep track of variable information. It really is more like a Doom-style hallucination than anything else. Some of the videos on that article are truly bizarre looking. I'd imagine after a few minutes every single one of them would devolve into an endless loop of being trapped in non-sensical geometry or killing the same enemy over and over again as the AI has no way of remembering the enemy existed to begin with, let alone that you killed it.
I'll be honest I don't think there is much use in this at all. It suffers from the same limits as any other model AI. Believability at a glance is not believability under scrutiny and if it's only believable at a glance then there's not much practical use in it. The advance in computational power and model sophistication required to stand up under scrutiny is massive.
The author seems to have written endless amounts of "AI" articles. Do they really not know what "hallucinate" means in such a context?
DOOM runs on everything, even on a hallucinations of artificial intelligence. It's getting beyond ridiculous.
- "Can artificial hallucination run Doom?"
- "What kind of stupid question is that??? Ofc it can, let me show you ..."
This is perfect for capitalism with Matrix bio-fuel-cells-human/battery tech!
It would have been too easy to just chill peacefully and unbothered in my cozy pod - they would feed me a hallucination of a dead-end job the whole time, complete with all the stupid office buttons I have to press.