/h/ai_slop House Furry Founder
DWHITE___________DYNAMITE DWHITED/YNAMITE :b: :u: :s: :s: :y: :space: :l: :m: :a: :o:  1mo ago (video post) 217 thread views #308054 spent 1,160 currency on pings

you can now matrix hallucinate losing at counterstrike in real-time with a stable diffusion neural network without actually sucking at it

https://diamond-wm.github.io/

!g*mers

20
New Old Top Bottom Controversial Random
Jump in the discussion.

No email address required.

View entire discussion

I dont get it

Jump in the discussion.

No email address required.

its just like the AI doom thing

they trained an engine to predict what CS looks like based on user input to whole cloth generate gameplay from scratch, no rendering or coding or anything, literally just a neural network of imaging trying to feed you what it thinks counterstrike should look like based on input and a marginal amount of preceding frames to maintain the instance

the real insanity is when you think about the intrinsic amount of math it takes to make a game like CS where as with AI you can simulate all of it purely based on training it with images essentially

Jump in the discussion.

No email address required.

Oh I see. Id imagine its actually easier for the AI to predict cs as the map (in this case) is always the same.

Jump in the discussion.

No email address required.

yeah for sure, its all clearly define geometry so theres a lot of predictability unless you maybe move up close to a wall for way too long and cause it to get scrambled

Jump in the discussion.

No email address required.

Link copied to clipboard
Action successful!
Error, please refresh the page and try again.