This post has been de-listed
It is no longer included in search results and normal feeds (front page, hot posts, subreddit posts, etc). It remains visible only via the author's post history.
Projectpage: https://diamond-wm.github.io/ Download and play it yourself: https://github.com/eloialonso/diamond/tree/csgo
It's likely this method or a successor method will be used as the renderer, not the entire game engine. The renderer only cares about what's happening on the current frame being rendered, along with some information from the last few frames. This would be a perfect fit for diffusion.
Games will just get bigger. Thanks to the advancement of game making technology indie studios are putting out games today that would have been huge headline games in the 90's (example: Sea Of Stars), but today they're lost in a sea of games. Studios got bigger, the technology got better, so games got bigger.
We can expect the same for AI driven development. If one person can use AI to create a game as big and intricate as Skyrim then a team of 500 people can make an even more massive game using the same AI tools. Skyrim sized games will become niche indie games.
When capacity grows usage grows to meet capacity.
Subreddit
Post Details
- Posted
- 4 weeks ago
- Reddit URL
- View post on reddit.com
- External URL
- v.redd.it/g9eik3y3oiud1
It will be interesting to see how context is solved. Just scaling it up won't work because a game can run for an arbitrary amount of time and I doubt gamers will be happy to have time limits come back. They'll have to find a different way to have the model remember what happened an arbitrary amount of time ago.