Welcome to the discussion of US Politics!
Rules:
Links must be to the original source, not an aggregator like Google Amp, MSN, or Yahoo.
Example:
We ask that the users report any comment or post that violate the rules and use critical thinking when reading, posting, and commenting. Users that repeatedly have comments or posts removed, weaponize reports, or violate the code of conduct, will be banned.
All posts and comments will be reviewed on a case-by-case basis. This means that some content that violates the rules may be allowed, while other content that does not violate the rules may be removed. The moderators retain the right to remove any content and ban users. The reason(s) behind moderator action(s) are publicly available through the Modlog.
That's all the rules!
Civic Links
• Congressional Awards Program
• Library of Congress Legislative Resources
• U.S. House of Representatives
Partnered Communities:
• News
AI can't hallucinate CSAM. It's trained. Maybe some people can skate by on first amendment, but the ones creating the models can't. Target the plague at the source.
If theres csam in the training data, surely they include where they got that image, and can use image recognition to find those images.
Saw this a few days ago, that Canada was reporting Grok having it in the databases of known children who were abused
I honestly don't think I'd be able to opt in to getting those notifications.
Doe’s images were included in a CSAM Hash List maintained by NCMEC, and “that same material” allegedly “was part of the dataset xAI used to build Grok’s image and video generating capabilities.”
You can't generate the original image from a hash, and it would make no sense for the original images to be publicly accessible, so I'm not entirely clear what the Canadian csam database has to do with this, unless the lawyer is saying it was leaked online any Grok picked it up from there or the article is saying images grok produced matched the hash associated with her images and alerted her.
Yeah I meant that Canada was somehow finding these images created and matching them back to images of children in their databases. Talk about a wretched job to have to exist. Whether the the images were in the original stuff to train the ai or used by a person say "make something off this image" and the ai allowing it.... All a gross situation.
The rumor is indeed that they got a hold of known csam and used it for training
It seems infinitely more likely that many pictures in the database can be found online or in other training sets than that xAI engineers secretly hacked the NCMEC csam database.
That is in fact what I meant when I said they got a hold of known pictures...
I mean it doesn't matter where the og csam images came from. The point is, they are saved in groks database and are being used to make ai csam.
And ncmec knows this because they match materials they have in their database.
It does matter because they have a more effective searching tool than ncmec, and should be alerting ncmec instead of adding it to training data.
Well. Yes. From that perspective it is important to learn how grok got the images.
It's possible they were shared through FB messenger but it's also possible they were gathered from other sources.
And those sources should be investigated and , like you say, why did grok find them when the ncmec hasn't ?
If it's from fb messenger then that explains a lot. But we don't know.
This sounds like an excuse to remove your rights. How is the government going to know what images people may have generated?
That's entirely unnecessary, they just need to remove csam from the training data.