Except if we're splitting hairs on this level, it's not actually that person, it's a likeness of that person.
... I don't think that has merit as a position, really, but I also don't think that "consensual porn" is accurate, either. I think we need a different term to involve porn made without the physical or mental involvement of an individual, to separate it from shit like sneaking upskirt shots or rape videos.
It's fun to lump them in together, and if you look just at the definition of the words... yes, it is. Sure. That's not really debatable.
But do you truly believe that drawing a distubingly-accurate porn picture of someone is equivalent to those other things? While the term may be "accurate", it's not helpful to distinguish the offense and its impact.
I think we should consider removing all identifying names from the corpi used to train these models. At the very least, I think someone should have the right to have their name removed from data labeling used in training.
Tangentially: It's always been possible for particularly-skilled artists to do a rendition of someone else that is very realistic and yet would ostensibly be without consent. They would do this by directly referencing real life photographs taken of an individual and interpolate details where missing, much like what's going on with our fancy parrot programs. Is this a fundamentally-distinct transgression, or are we only concerned about it because of how easy AI has made this transgression?
Not sure there's really a discussion about equivalency to be made here - what it is is wrong inherently. A human doing it by drawing or meticulously rendering each frame of a digital video would be equally wrong to an AI, but now AI allows anyone do it without difficulty as you mention in your last paragraph.
Agreed. It's wrong. Pretty sure I didn't say it was alright in any circumstance, but if you can hilight where I did, I'll correct it from my post, as that wasn't the message I wanted to send.
There are degrees of wrongness.
Drawing a lewd picture of someone is not rape, for instance, but both are wrong and both have been mentioned.
We have separate terms to describe different things for a reason. That was the point I was trying to get across.
It's just a shame IMO that leaving out the word AI-generated (or deepfake) hurts the point that the article wants to make.
This is an example of the negative effects of binary or black/white thinking, IMO (not sure what's the correct term in English)
Parent comment did not want to downplay the seriousness, personal harm and moral+judicial implications of realistic fake porn, as I read it.
The headline in question just weakens the public attention for actual rape porn depicting physical rape, or nonconsensually filmed sex for anybody reading the article, I'm afraid.
Especially when celebrities come into play.
It's true that both crimes can converge into something very similar, but these are still distinct crimes (maybe even happening in combination!?) which need to be examined thoroughly.
Escalating language in headlines doesn't achieve its ostensive aim here, IMO.
There have been discussions about deepfakes and their dangers, but not enough of them, especially when it comes to non-celebrity persons.
I'm afraid escalating language won't help to catch more attention from people who previously underestimated the severity of current and potential dangers.
I see the sides of the argument but the argument can go on and on until someone is willing to step up and decide whether an image of a person is a person themselves. Otherwise it’s unresolvable.
... I don't think that has merit as a position, really, but I also don't think that "consensual porn" is accurate, either. I think we need a different term to involve porn made without the physical or mental involvement of an individual, to separate it from shit like sneaking upskirt shots or rape videos.
It's fun to lump them in together, and if you look just at the definition of the words... yes, it is. Sure. That's not really debatable.
But do you truly believe that drawing a distubingly-accurate porn picture of someone is equivalent to those other things? While the term may be "accurate", it's not helpful to distinguish the offense and its impact.
I think we should consider removing all identifying names from the corpi used to train these models. At the very least, I think someone should have the right to have their name removed from data labeling used in training.
Tangentially: It's always been possible for particularly-skilled artists to do a rendition of someone else that is very realistic and yet would ostensibly be without consent. They would do this by directly referencing real life photographs taken of an individual and interpolate details where missing, much like what's going on with our fancy parrot programs. Is this a fundamentally-distinct transgression, or are we only concerned about it because of how easy AI has made this transgression?