SquishyPillow

joined 1 year ago
[–] SquishyPillow@burggit.moe 4 points 1 year ago (2 children)

λ=749nm. Hardly anyone would notice it.

[–] SquishyPillow@burggit.moe 1 points 1 year ago

Are you inpainting the eyes at all? Eyes often get screwed up due to autoencoder artifacts. Some models that use a custom autoencoder like AnythingV3,4,5,etc. don't have this problem, probably because their training data contained lots of images of people with large eyes. You could try training a custom autoencoder, or use a model with one.

Or just inpaint the eyes, and make sure that the inpaint area is set to "only masked".

 

All 50.

At the same time.

In lockstep.

Of course. lol

[–] SquishyPillow@burggit.moe 1 points 1 year ago

It wasn't a warning, it was a vision. Look up who the Huxley family really are.

[–] SquishyPillow@burggit.moe 1 points 1 year ago

hi fwend <3

i will post what i can. but mine will mostly be SFW hehe

[–] SquishyPillow@burggit.moe 2 points 1 year ago (1 children)

lolis filter moralfags out

[–] SquishyPillow@burggit.moe 2 points 1 year ago (1 children)

maybe the "pantyhose" keyword is too correlated with pictures of mature ladies, so thats why it happened. thats all im gonna discuss this though.

[–] SquishyPillow@burggit.moe 5 points 1 year ago (3 children)

they have big squishy bums i think

[–] SquishyPillow@burggit.moe 2 points 1 year ago (1 children)

Even if they do that, it will be futile. Here is an explanation of why.

[–] SquishyPillow@burggit.moe 2 points 1 year ago (3 children)

So far new reddit seems easier to scrape than I initially thought c:

[–] SquishyPillow@burggit.moe 1 points 1 year ago

Yeah :c

A good reminder for me to reverse-engineer new reddit's backend once this happens.

 

Made this as a replacement for URS as soon as Reddit announced their API changes c:

 

TL;DR: The existing incentive structure makes it impossible to prevent the creation of large machine learning models like Yudkowsky and others want.

Also, keep in mind that the paperclip maximizer scenario is completely hypothetical.

 

TL;DR: AI will be able to generate more data after it has been trained that is comparable to the quality of the training data, thereby rendering any training data absolutely worthless. The time to sell data at a reasonable price is now, and those locking their data behind huge financial barriers (such as Twitter and Reddit) are stupidly HODLing a rapidly deprecating asset.

 

TL;DW: Blackmail is dead because generative AI will eventually be so realistic that anybody will be able to use it for plausible deniability. Anyone currently blackmailed will be freed once this happens.

This explains why all the AI companies are desperately trying to censor/align their models.

view more: next ›