[SXX]

We can speculate all we want. The fact is: its impossible to apply any real restrictionsn ML model. Just like any DRM this one gonna be easy to bypass and will only degrade experience of legit users and gonna slow down technology advancement.

We dont require image editors to have CSAM filters and requiring filter for ML model is madness.

[hackerlight]

You're talking about filters inside the ML model, but I think the more interesting question is morality and the law.

[yellowapple]

"The law" is kind of irrelevant unless it's enforced, so it really boils down to morality.

There are two separate questions to be answered here:

1. Is it immoral?

2. Is it sufficiently immoral to warrant imposing my own moral standard upon others, by force if necessary? i.e. is responding to or preventing it with violence itself moral?

Morality is subjective and relative, which is why it's interesting to ask both questions: given the subjectivity and relativity of morality, at what point does your own moral code take precedence over those of others?

[candiodari]

If there is no technical means to enforce the law, there is no law. At least when it comes to the war on drugs ... at least there's some physical thing being traded along. These models would generate such imagery and then immediately destroy it.

This will be a bigger disaster than the war on drugs, because it's almost impossible to provide proof. So what's left is convicting people based on oral testimony alone ...

[SXX]

I am personally on conservative side here so I'm not questioning morality and the law we have now. I'm very much against of spreading of such material regardless of how it was created. If someome decide to publicly share something like that there should be consequences.

At the same time I think everyone should be free to wrire / paint / generate whatever material they want on their own computer within comfort of their house. As long as no one gets hurt I dont give a damn.

[guilhas]

> The fact is: its impossible to apply any real restrictions ML model

What? Of course you can curate what you feed and have filters to train your ML model

I am sure it is currently easier to train a model with legal sex content and without any children

Also the law is still the same, having and distributing pedophilia content is illegal

[SXX]

This is not how ML works. You don't need real child abuse material in your training set to make a model capable of creating CSAM. All you need is depictions of any children and any sexual content.

[guilhas]

That is in line with what I said

Most publicly available sources of sexual content are adult, legal, don't have children

So if you feed Pornhub to your model, which most porn is actually well tagged, as is many ML data, you really have to go out of your way to get any children

Not saying is not possible, I am saying is easy to avoid. Not as an impossibility to separate both as you suggested above

[SXX]

Nope. You missed the point. I said the opposite.

If your training data have any absolutely innocent children pictures at all and also NSFW content with adults then model will be able to generate NSFW content with children. It's that simple.

Also even if you remove any children completely from training model will still be able to generate CSAM-like content because a lot of legit porn include jailbait models.