Winterland getaway
Added 2024-11-11 01:14:12 +0000 UTC
Comments
I dont have the technical knowledge of the underlying stuff but yeah it all seems to imply the only thing stopping generative tech from making something we'd see as actually alive is just the ability to perceive more of physical reality and string together more disparate parts to kind of make it a cohesive concept to interact with. Maybe, we'll see I guess
futawarp
2024-11-11 21:33:24 +0000 UTCRamble away whenever you feel like! It's an interesting comparison between natural/human learning processes and a generative algorithm consuming everything given to it. The former gradually learns context-sensitive associations with a semantic layer on top through specific examples while latter models extremely loose, context-free associations with limited semantic pairings. With enough high quality data you can simulate natural learning with a model, but such data is often expensive in sufficient quantities so the easy road is taken i.e. throw everything at the algorithm and try to tweak the model as you go. This is the state the generative AI industry at large is in, whether it be Stable Diffusion, ChatGPT, etc. It's exciting to think of their potential for improvement in all manner of contexts. :D
Tempre
2024-11-11 08:11:33 +0000 UTCLoving the muscular futa plowing away 🍆💦🍆💦🏆🏆🏆🏆🏆
Jakes steed
2024-11-11 06:46:54 +0000 UTCSorry for the ramble and thanks!
futawarp
2024-11-11 03:24:00 +0000 UTCThat whole thing is interesting to me, the slightly incoherent way these things blend reality. People talk about dream logic. I don't know if it works 1 to 1 but I like thinking of each model like a brain innured in the concepts of color and the correlation to words without context. No context for the places created between those correlations and even no context for the words or colors. Just a mind submersed in one thing, simulating perhaps a lifetime of of understanding of this one hyper-specific thing. Our brains have thousands of different connections acting constantly which adds context and a fuller understanding but makes doing such hyper-specific tasks difficult without similar training.
futawarp
2024-11-11 03:21:32 +0000 UTCI'm not into the horsecock, but I like how detailed the environments turned out in this batch! There are some logical inconsistencies (door attached to a window, I'm looking at you), but overall the model is producing some nice output here. Keep up the great work!
Tempre
2024-11-11 03:01:29 +0000 UTCThis set is superb, keep up the great works🔥🔥
BACONAD3R
2024-11-11 02:09:14 +0000 UTCAlways happy to see azula and chel
Anything
2024-11-11 01:16:40 +0000 UTC
