[ - ] prototype 1 point 10 monthsJun 22, 2023 03:43:50 ago (+1/-0)
If they were going to identifty memers they would have done it years ago, like so
1. release tons of free software and promote it. embed the code for identification in free, popular, 'open source' programs. in practice its been shown that very few code audits happen, even on major open source projects. and even fewwer people understand the whole codebase so bad faith changes included prior to release, are rarely caught
2. compiler level invisible trojans, which are officially something that exists not just in theory, but now in practice, compromising potentially everything
3. hash-invisible exploits, by releasing public hashing standards that contain meta-collisions (collisions that are invariant to very specific data inclusions in a larger set)
4. os level compromises
5. hardware compromises.
I know of examples for four out of five of these. This is just them announcing the programs that already exist worldwide.
Stable Diffusion projects are open source. The article mentions using a plugin to add the watermark. Don't use the plugin; no watermark. Any attempt to add this to the base code can and will be swiftly and easily reversed.
Closed source, proprietary AI apps... that's a different story. This or something like it will inevitably be embedded in all of them, and removing it would be next to impossible. So don't use them.
[ + ] FacelessOne
[ - ] FacelessOne 2 points 10 monthsJun 22, 2023 03:30:46 ago (+2/-0)
[ + ] prototype
[ - ] prototype 1 point 10 monthsJun 22, 2023 03:43:50 ago (+1/-0)
1. release tons of free software and promote it. embed the code for identification in free, popular, 'open source' programs. in practice its been shown that very few code audits happen, even on major open source projects. and even fewwer people understand the whole codebase so bad faith changes included prior to release, are rarely caught
2. compiler level invisible trojans, which are officially something that exists not just in theory, but now in practice, compromising potentially everything
3. hash-invisible exploits, by releasing public hashing standards that contain meta-collisions (collisions that are invariant to very specific data inclusions in a larger set)
4. os level compromises
5. hardware compromises.
I know of examples for four out of five of these.
This is just them announcing the programs that already exist worldwide.
[ + ] GloryBeckons
[ - ] GloryBeckons 0 points 10 monthsJun 22, 2023 05:15:07 ago (+0/-0)
Closed source, proprietary AI apps... that's a different story. This or something like it will inevitably be embedded in all of them, and removing it would be next to impossible. So don't use them.