remember...
this post was submitted on 13 Apr 2025
882 points (92.1% liked)
50501
1576 readers
694 users here now
50 States, 50 Protests, 1 Movement. https://fiftyfifty.one | #fiftyfiftyone
This is a community related to the 50501 movement. Find the main communities at https://50501.chat.
Rules
- NO VIOLENCE: This is a non-violent political movement! We do not tolerate violence or calls to violence; if that's what you are interested in, you're in the wrong place. Any comments or posts discussing firearms, weapons, violence, or any form of incitement will result in an immediate and nonnegotiable ban.
- RESPECT: Be respectful. Trolling, strawmanning, abuse, etc. will not be tolerated.
- INCLUSION: This movement is for everyone. Exclusionary language will not be tolerated.
- DOXXING/HARASSMENT: Doxxing or harassment will not be tolerated. Instant ban for violations of this rule.
- NO SELF-PROMOTION
founded 3 months ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
That doesn’t work, based on the results we’ve seen so far. All attempts to poison AI have quickly been bypassed.
Is there any source to this claim?
Sure, a quick google search turned up this, published just four days ago:
https://ieeexplore.ieee.org/document/10962550
Or here’s a discussion on r*ddit about how Glaze is totally ineffective against poisoning image generation AIs. (Same principles apply to others like Nightshade)
https://www.reddit.com/r/aiwars/comments/12f9otc/so_the_whole_entire_glaze_ai_thing_does_it/
Basically every poisoning technique is by its very nature too specific to be useful for poisoning all the types of AI out there, or is too invasive and ruins the content for humans too, or is not strong enough and it gets lost in the cacaphony of data anyway.
Really it’s a clever idea in principle but in practice AI poisoning isn’t really a mathematically viable way to combat AI.