this post was submitted on 21 May 2024
510 points (95.4% liked)
Technology
57350 readers
6207 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Does this mean the AI was trained on CP material? How else would it know how to do this?
It would not need to be trained on CP. It would just need to know what human bodies can look like and what sex is.
AIs usually try not to allow certain content to be produced, but it seems people are always finding ways to work around those safeguards.
Local model go brrrrrr
Training is how it knows it...
You can ask it to make an image of a man made of pizza. That doesn't mean it was trained on images of that.
But it means that it was trained on people and on pizza. If it can produce CSAM, it means it had access to pictures of naked minors. Even if it wasn't in a sexual context.
Minors are people. It knows what clothed people of all ages look like. It also knows what naked adults look like. The whole point of AI is that it can fill in the gaps and create something it wasn't trained on. Naked + child is just a simple equation for it to solve
You can always tell when someone has no clue about AI but has read online about it.
The whole point of those generative models that they are very good at blending different styles and concepts together to create coherent images. They're also really good at editing images to add or remove entire objects.
I think @deathbird@mander.xyz meant was the AI could be trained on what sex is and what children are at different points. Then a user request could put those two concepts together.
But as the replies I got show, there were multiple ways this could have got accomplished. All I know is AI needs to go to jail.
Likely yes, and even commercial models have an issue with CSAM leaking into their datasets. The scummiest of all of them likelyget one offline model, then add their collection of CSAM to it.
Well some llm have been caught wirh cp in their training data