for context: https://www.theguardian.com/environment/ng-interactive/2026/oct/22/finland-google-felling-trees-datacentre
“‘Pure arrogance’: anger in Finland as Google illegally fells 250,000 trees for datacentre”
for context: https://www.theguardian.com/environment/ng-interactive/2026/oct/22/finland-google-felling-trees-datacentre
“‘Pure arrogance’: anger in Finland as Google illegally fells 250,000 trees for datacentre”
I ran it though and AI image detector - zerogpt. They can scan for invisible water marks that AI places on images.
Yeah but for that’d they’d have to have access to some pretty damn valuable trade secrets, so its plausible the website is lying
Honestly, that’s how I was told that they work, and looked into it with a reasonable degree of due diligence. I’m not technically strong on computer stuff but from what I read that’s how it works.
I don’t get Lemmy’s AI stance in general - half the folks want AI called out and bust out the pitchforks at the slightest hint of it, and others like you call AI detection shit, which is also fair.
Instead of being a drooling AI policeman behind a keyboard, typing out ‘SLOP!!!’ next to a jarred Rainbow Dash figure, however, I added some context about the article in my original comment.
I don’t know what you guys want anymore lol
Thats valid, and I feel like it’s easy to expect everyone here to be fairly read up on the topic, when really only a vocal minority actually exists lol.
I’m no AI researcher, but I’ll admit AI is cool on a mathematical level so I’ve read a lot. If you’re interested, I can kinda try to describe how I’m anti AI, and AI detector.
One of the somewhat recent AI breakthroughs is the scalability of generative adversial models. All that being fancy talk for we like 2 robots arguing against each other, because it works well on a calculator but also on a supercomputer.
That scalability caused all the hype and stuff, but importantly (at least image models) are 2 models trained using super computers to detect/fool each other. This means that, to make a detector, any small company would need half that equation - but it would have to be slightly better than the LLM companies supercomputers. And even then it’d be just to detect a signal slightly more often than noise (so real expensive to probably not even work).
Even more recently, this ‘unable to detect without a supercomputer’ became dangerous for regurgitating training data, poisoning models with themselves, so a lot of watermarking techniques were developed.
Unfortunately the companies making these watermarks aren’t idiots. They do it similarly to cryptographically signing a document, but hide a signal within the data, and keep the key to themselves.
Without that key, any detector is essentially trying to rebuild encrypted data that they don’t even know where is, which is to say entirely useless.
Ugh, so the detectors are just AI as well? And you can’t even check accurately without a super computer?
I should just throw in the fucking towel with even trying to sleuth it out. I post a lot. A lot a lot. Because of the volume some of my posts end up getting flagged as AI. I wasn’t able to check really until about a month ago and now you’re saying it doesn’t even work to begin with.
Can I get permission to kick the next person that posts the stupid slop cat in their gonads? Im tired boss…