The irony of course is the reality of how the models work.
It's trained on an obviously very safe, very kosher, very approved dataset, with all the even remotely outstanding generations shot out of it, and the fact it is very good at generating female bodies in bikinis at every conceivable angle tells us... What?
Are whores ready to admit slutposting detracts from all women?
These models are trained on everything, including tons of porn. It's not the training data that determines their outputs. It's the guardrails the engineers place on the model to both curate user inputs and safety check model outputs.
The irony of course is the reality of how the models work.
It's trained on an obviously very safe, very kosher, very approved dataset, with all the even remotely outstanding generations shot out of it, and the fact it is very good at generating female bodies in bikinis at every conceivable angle tells us... What?
Are whores ready to admit slutposting detracts from all women?
These models are trained on everything, including tons of porn. It's not the training data that determines their outputs. It's the guardrails the engineers place on the model to both curate user inputs and safety check model outputs.