To qualify, a video game must have its graphics rendered by AI in real time. An AI should be rendering graphics at least 50% of the time in an average playthrough. When the AI is rendering graphics, it should generate at least one image per second. The images should generally depend on the very recent actions of the player - for instance, the player character immediately starts moving left when you press the left button.
It's okay if certain parts of the game are not AI-rendered, like UI elements. However, it doesn't count if the AI is just starting with e.g. a 3D mesh rendering and postprocessing it to make it prettier or fill in frames. It has to generate the video more-or-less from scratch. Specifically, I'd look for a system that generates visual elements like characters and scenery during gameplay and simulates their motion, doing this entirely through ML rather than programmatically, similar to OpenAI's Sora. It's okay if the AI starts with a premade textual or visual prompt, but has to be able to generate and animate new visual elements.
The following is speculation, not resolution criteria: What I'm imagining is a very open-ended game where the AI makes up new landscapes and encounters for you on the fly, sort of like the Mind Game in the book "Ender's Game." This seems doable by a very fast version of OpenAI's Sora, where you just ask it to generate video game footage on-the-fly and conditional on the player's button presses.
I may bet in this market.
Markets with the same resolution criteria
/CDBiddulph/will-a-video-game-with-airendered-g-3b1136cb8e7c
/CDBiddulph/will-a-video-game-with-airendered-g-30dbd8e5eede
/CDBiddulph/will-a-video-game-with-airendered-g-2e37a190fefe (this market)
1,000
3.00@MalachiteEagle I think it's plausible that 1 million people have played it on their web demo by now
@MalachiteEagle Yeah, I would count it if 1 million people downloaded a game like this rather than just playing it in their web browsers! I think most games do get downloaded, so hopefully the web browser is not going to become the platform du jour for AI-generated games or I'm going to have some trouble with the resolution criteria
@CDBiddulph well, when you connect to a website, the browser does download the content. So playing this game in a web browser qualifies as downloading it. You might want to explicitly exclude web browsers in the description if that's not what you had in mind.
@CDBiddulph That being said, the resolution of this question may also get confusing if native games with AI-generated graphics offload some of the neural network processing to the cloud.
@MalachiteEagle Yeah, I expect the neural network processing will likely be offloaded to the cloud. I guess what also makes me feel like this shouldn't count is that no one is paying for this game and they can only play it for 5 minutes, but admittedly that is not part of the resolution criteria
@MalachiteEagle Interesting. I'm still leaning towards no, but now that there's actually confirmation of the number of users, I feel like I could be swayed if enough people support the market being closed. Please like this comment if you agree the market should be resolved, or comment if you think it shouldn't be
I will say, all 4 of my markets on this topic are probably underpriced at this point
https://arxiv.org/pdf/2311.18828 This new technique distills an image generator to let it create 20 high-quality images per second. I don't know if it would extend to a video generator like Sora, but it seems likely. It's a very small step from a fast video generator to an AI-rendered video game - you just have to add controls.
Something that would qualify and seems the most likely to me, actually I have already seen versions of it, is a simple interactive story telling with an image being rendered after each decision. And to satisfy the 1 frame per second constraint it would just animate those images in some simple repetitive way.
@patrik I think that probably wouldn't count, since "the images should generally depend on the very recent actions of the player." If the player takes 10 seconds between decisions on average, even assuming the first image after each decision gets rendered quickly after the decision was made, only 10% of images would "depend on the very recent actions of the player."
@patrik Not necessarily, if the game doesn't let them input their next decision immediately. The point of that part of the resolution criteria was that the game should mostly be rendered in real-time (like a platformer or a shooter), not periodically rendered (like a visual novel or some turn-based RPGs).
New paper on "Genie" from DeepMind: https://sites.google.com/view/genie-2024
As far as I can tell, Genie already qualifies to resolve this market! ...Except for the key detail that the game must get 1 million downloads, and it's currently way too boring and slow to get anywhere close. But this is a significant step forward, coming just a week after I posted the original market! It even generates frames at 1 FPS (the bare minimum for this market), according to one of the authors on Twitter
Genie could enable a large amount of people to generate their own game-like experiences. This could be positive for those who wish to express their creativity in a new way, for example children who could design and step into their own imagined worlds. We also recognize that with significant advances, it will be critical to explore the possibilities of using this technology to amplify existing human game generation and creativity—and empowering relevant industries to utilize Genie to enable their next generation of playable world development.
@CDBiddulph Genie is more like a research project, but yes it def qualifies. I would probably expect that we might see a game like that would get a mil downloads because people would be interested in a game that is fully made using AI even if it is a 2D game
@MichaelWheatley Right, that wouldn't count. There's some discussion of this in the comments of https://manifold.markets/CDBiddulph/will-a-video-game-with-airendered-g