this post was submitted on 28 Aug 2023
1736 points (97.9% liked)
Lemmy.World Announcements
29163 readers
36 users here now
This Community is intended for posts about the Lemmy.world server by the admins.
Follow us for server news π
Outages π₯
https://status.lemmy.world/
For support with issues at Lemmy.world, go to the Lemmy.world Support community.
Support e-mail
Any support requests are best sent to info@lemmy.world e-mail.
Report contact
- DM https://lemmy.world/u/lwreport
- Email report@lemmy.world (PGP Supported)
Donations π
If you would like to make a donation to support the cost of running this platform, please do so at the following donation URLs.
If you can, please use / switch to Ko-Fi, it has the lowest fees for us
Join the team
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Isn't there semi-automated tools that can detect CP?
Those might be an automated solution to at least cut down on the volume.
The same can go for banned images. These can automatically be identified with perceptual hashing, and automatically be denied when uploading.
We should just track these people and send them the police directly, way more efficient than Symptom control.
There are. They are easy to abuse too, and cause a lot of other issues, along with not being 100% efficient. Apple tried that and was met with appropriate backlash. There are many issues : training data is one, but also it is impossible to rule out false positive/false negative automatically, and it is relatively easy (for now) to doctor pictures to pass through. A nefarious actor could easily bypass these. There's also the case of false positive that can very quickly push some bystander under the bus, and knowing how moderate and understanding internet is⦠yup.
It is also risky, as this will effectively be a censorship tool; they are often setup in the pretense of "helping", but once they're up, it all depends on who steers it. Such responsibility can hardly fall on moderators/admin of lemmy, and it would also be problematic to handle them on a more nationwide (or more) level, since it would give incredible censorship power to authorities.
And a bigger bottom-line is that working hard to prevent these kind of content from reaching us, while it have an obvious upside, also does nothing in respect to the actual issue of the content existing and being created.
tl;dr: there is no easy solution that doesn't come with one hell of a string attached to it.
On the other hand, it is quite hard to hide yourself from authorities online, and this kind of behavior (I hope, somehow, that the people that posted these only do so to be toxic to lemmy and not to actually disseminate content) should lead to some action from authorities, getting to actual people and subsequently moving upward the stream to actually act on it. Hopefully.
The one thing AI would be good for... Humans shouldn't have to see that shit.
Humans have to see that shit to train ai. That is why it is so difficult to find a model for that
Honestly what a shit job you must have when working with that stuff. it is sad and a difficult thing.
A lot of the mods for big providers like FB require counseling after the horrible crap they see (not just CSAM, but also terrible things like animal abuse and mutilation, etc). Unfortunately, the big companies have outsourced much of it to other countries where there aren't as many worker protections, traumatizing people and replacing them when they can't meet some arbitrary metric.
You'd need data to train an AI... Yeah that won't happen
IIRC there is a database that law enforcement uses during investigations to obtain access to these groups (they obtain consent from the victims to use the material).
the one thing you do not do with this shit is use AI (too much risk of false positives. I mean, I remember Facebook's AI thought that an onion was nudity (probably thought it was a boob).
But tools do exist. PhotoDNA by Microsoft. Although much more user-friendly implementation if you use Cloudflare, related links:
the "risk" of false positives comes down to the consequence. if the consequence is being stuck in the slammer, don't use ai. if the consequence is you can't upload the image unless you manually appeal, or even maybe have to use an external image host; i think ai is fine
edit: ah bugger, wrong acct. ah well
(please tag @zeus@lemm.ee if you want me to see your response)
Apple is building the tech, too they arent sharing it with the world for humanitarian purposes like this.
PhotoDNA.
Apple willingly made that tech, however its designed to get the individual in trouble by looking at the photos from their personal device and snitching on its owner (leaking them in the process). They can easly adapt it into a tool for social media, but it goes against their beleafs.