No child is being abused when someone play with weights on a model.
"Abhorrent nature" is very subjective and relative.
There’s also a limit to going too far in connecting abstractions.
I think the parent was making the point that creating and distributing simulated child pornography should be. The medium that is used to reproduce the image is irrelevant. Real digital images are also made of 1s and 0s.
Even in the extreme case, where it really does contain images in binary format, there is only the latent potential of that being transformed into something objectionable, several steps removed.
If this was the accepted norm everywhere, then practically every adult in the US would be guilty of some latent potential crime, after several degrees of separation.
e.g. even though lunatics use knives in mass stabbing sprees, we don’t assign blame to those selling knives or knife making equipment in stores where the lunatic happened to have gone shopping.
If we are talking about a physical book containing 1s and 0s that represent CSAM, I'd ask how you encoded the image to get the binary representation...
Intermediaries who likely cannot recognize it and won’t think twice about reproducing it, even if later on it does get transformed back.
What the original impetus for creating and distributing it is irrelevant to anyone subsequently encountering it.
To continue the knife analogy, even if the knife factory was also in fact owned by some maniac, the salespersons in the retail stores still not assigned blame even if they promote the knives.
The salespersons are just random middle men who couldn’t possibly have been expected to personally investigate the actual background or provenance behind what they’re handling.
You could even stretch the idea further to say that someone making a bomb is merely mixing certain perfectly legal chemicals just so...
1s and 0s can be sending bitcoin to a hit man to kill your wife. In this case, link to harm is indeed obvious, and evil and years in jail are justified. Can you argue for something similar for generated CP? And also discount the ick factor?
I don't know if theres a correlation between evil and jailtime. The debate around the severity of punishment is different than the debate about whether or not something should be illegal. I do think my (US) justice system is very flawed, especially around sentencing.
> Can you argue for something similar for generated CP? And also discount the ick factor?
I can take a stab at it. Simulated CSAM :
* if realistic, can make identifying real CSAM and the associated victims more difficult. This could redirect resources dedicated to stopping harm against victims.
* has the potential to normalize the distribution of CSAM due to #1
* has the potential to normalize the consumption of CSAM.
To address the comments being birthed around unproven causal relationships- yeah I know. The parent asked for arguments. Besides, a study into this doesn't seem very ethical or possible.
I am not able to discount the "ick factor", but I'll try to make a counterpoint.
All of the arguments I can think of to the contrary center around free speech, false positives and weaponization. That's where I think the challenges are and where subjectivity can cause issues. Without a real person, it's difficult to assign a simulated age unless a prompt is captured or other context. Otherwise it becomes the job of a person or other 'AI' to guess the simulated age of the image.
Yes, but at the same time it'll greatly devaluate the real CSAM. It'll be like flooding the black market of perfect copies of rhinoceros horns. People won't have any economic incentive to create real CSAM.
No matter how abhorrent a picture is it can not harm anybody by merely existing. Harm can (or may not) only be inflicted to others during its production or with its demonstration.
Banning possession of pictures produced without participation of any living person other than the viewer makes no sense.
Banning possession of child abuse pictures used to make sense because possessing implied buying and buying incentivized production which depended on actual child abuse. Nowadays nobody has to be abused and I'm totally fine with weirdos using whatever pictures their sick imagination may want as long as this keeps them satisfied enough to be able to keep their wicked desires secret from every living person. I don't want the police to bust such people as long as they don't actually harm anybody and don't pay anybody to do the harm. No crime without an actual victim is worth ruining anybody's life, whatever kind of freak they are.
It's the same logic most countries use to ban possession of firearms. Their presence greatly increases the risk of lethal violence and suicide, despite being harmless to people when used correctly at the range or for hunting. The lack of need for firearms weighed against the harm they cause to others makes it a hard sell for remaining legal.
Firearms may get a pass because they have real utility in self defense, but there are no positive benefits to having an endless stream of fake child abuse.
For a more than just relevant part of people who would play with weights on a model to create anything that compensates for "child abuse" with "fake child abuse", the risk of lowering the threshold is extremely high. VR, AR, fake news and the scripted nature of some world events that are driven by common subgoals of interest groups are decrementing the opaqueness of the line between reality and delusion/illusion. Loss of reality & dissociation from humanity are things, especially among people with a fanbase & influence, money, power.
Games & TV and their influence are obvious and they do erode the threshold as well but it's reversible because the player is not the creator of his reality while with gen ai, prompts & weights, the player becomes more and more capable of simulating his own reality based on his derived preferences & idiosyncrasies. Games and TV have constraints that are set by the devs and show runners and all the crews involved in the process. Gen AI doesn't have that.
I have not tried to reduce my hypothesis but there's plenty of discussions and research into mental health crisis, recurring and one-time/impulsive (sexually) violent behavior as well as the correlating &or causal psycho-pathologies.
But I'm not a scientist. I just think. And sometimes I read.
It matters anyway. It's hard studying this stuff in the lab. Relevant measurements in the field need to be isolated but that can mostly only be done in retrospect and we are talking about LLMs, what can be generated, what will be generated and should it be legal or not?
Should the system prepare? Can it? What kind of colony will we become? Are there multiple endings/"liminals"/paths? Who (exactly) cares? Who will be affected? Who is aware and who isn't? Who is misguided (maybe it's me)?
If I was a different kind of idiot in the past, I would have made the money to have the peace of mind to dive deeper but as it stands, just babbling some thoughts isn't gonna get me anywhere. Except that it stimulates to put it into words and maybe get something coherent on paper some day, something that isn't purely wannabe-megalomaniac, but constructive ^^
That statement is carrying a lot of non-evidence with it.
Say my fetish is chocolate. If I ask an image generating AI to give me a photo of an adult made of chocolate and covered in sprinkles, that doesn't necessarily mean it was trained on that specific thing.
You don't seem to understand that "potential" is not "actual" (and also conveniently says nothing about actual probabilities). I could claim almost anything is "potential".
Eating grapes could "potentially" result in choking. Living in a house made of wood could "potentially" result in dying in a house fire. Driving over 70MPH could "potentially" result in hydroplaning if it is raining.
Why do we not care more about the "actual" harm (and its probability), as measured by science?
Oh, of course- Things that are banned are difficult to study, which is why people can invent any beliefs about them that they want to.
I've watched my own three children learn to talk, then to write. The human mind is also trained on inputs.