neat will check this out, havenāt really watched any Super Mario World/credit warp runs before and looks interesting.
yep. allows you to get to the credits without ever leaving the first level
yep that a sure way to break the world record, could do the game in like 5 minutes.
current record is sub 55 seconds
damn thats quick, didnāt expect itāll be as fast as that
damn thats amazing. is interesting seeing the different strategies that developed over time to execute credit warps.
hmm for some reason Fedora refuses to install the bootloader on my laptop. have tried it with both standard Fedora Workstation and a netinst image and it fails each time and checksums match.
sounds like a bad drive
yeah is what Iām thinking. its a really old Intel ssd and iirc had some issues with it before.
That video sent me down a rabbit hole of neural networks playing video games. Now I want to train a neural network to play Forza⦠Although, thatās probably unrealistic.
would be neat to create a neural net that can play SMW. the main problem with most of the neural net works that play games is that they dont provide enough info to the net. only basic info about where blocks are and not what type of blocks it is
Well impulse Purchase
havenāt really looked into neural nets much but donāt most just aimlessly play through the level with not much context at all other than death and to avoid it until it eventually knows enough to complete it?
yes and no. Most just tell it a basic layout of the level and train the net for that level and just bruteforces it way through it until it finds a pattern that works
Isnāt that sorta cheating because the neutral net has pretext about the level and isnāt really learning anything other than a pattern?
still yes and no. it has a limited scope
Well, donāt speed-runners do the same thing? Iām no good at games, so Iām just making assumptions here, but I would hazard a guess thatās what a lot of them do.