[posting_mess]

Tangent, but related vaguely;

If you can "picture a brown cow" in your mind can you picture "the unholy" in your mind?

It seems logical that there is no universal constraint preventing anyone capable of picturing a brown cow from picturing the unholy, they just choose not to (or in some cases choose to).

I guess as shown, restricting ML/LLM/AI pathways after the fact has a negative effect on intelligence.

So I ask could you be word played into supporting the unholy by a good "sales person"? If you can are you intelligent a toll? What if you needed to for Science or safe guarding?

"Is context enough and whats the context of the context" I guess im asking.

The content depicted in the article is of course abhorrent. But how do you go about negating it when any intelligent being is likely capable of generating it internally?

[pavel_lishin]

There's a significant difference between imagining CSAM in your head, and having it displayed on a monitor in front of you.

If there weren't, movies, television, radio, theater & books wouldn't exist, since everyone would be rotating their own cow in their head for free.

[posting_mess]

A more concise version of my question is;

"How do we train large ML/AI systems to think generating the unholy is bad without hurting intelligence, given we know that applying some universal law (I.E RLHF) hurts the model".

Trying to promote the exact opposite of "let them eat the unholy".

[k__]

1. Many people aren't able to generating visuals in their mind.

2. We can't yet reliably extract the images from our mind and share them with others.

[SR2Z]

> 1. Many people aren't able to generating visuals in their mind.

Yes, but "picture a..." in this context is not specifically meant to talk about visuals. It means "recall the nature of..." and is a multisensory experience that is required for anyone to use language.

The point here is that if you have a word, it refers to SOMETHING and porn-sensitive companies don't always want an LLM to recall it.