on /r9k/ you can't post content that's been posted before. Duplicate images or words aren't able to be posted. This means that there is a database out there with every post ever made on /r9k/ right?There's a concept known as digital decay, which refers to online media just kind of disappearing for a variety of reasons. Pew recently did a study showing that about 40% of websites since about 2013 have just vanished. They are inaccessible. There's a maze of dead links that go nowhere. So there's an idea that, oh, if I put it out there, it will be online forever. I can always find it again. An employer can always find it. And that's just really not the case, especially now that we're in this kind of app-based world where you don't control your content. But /r9k/ is different...and vulnerable
>>106573244>what is a hash
>>106573256Something that you smoke?But seriously, it never works for prevention because it's infinitely easy to slightly modify an image and then spam it. Surely there's some AI that could be developed to prevent stuff like that? One solid usecase for AI and nobody's working on it! As far as I know anyway.
>>106573244>This means that there is a database out there with every post ever made on /r9k/ right?why don't you just look inside the leaked source code instead of asking dumb questions? it's all there
>>106573291By God this post triggers me to death. If only it were just a random idiot on 4chan, but people who think like this really walk among us. And worst of all are listened to.You have a technical problem that is most likely already solved by some clever algorithm painstakingly developed and published by people who put in effort and thought. If not, this is exactly the kind of problem begging for a computational solution.Instead of a web search or a little thought, in barges the hero of our time. He says the magic letters.>"AI"Genious. Why didn't anyone think of it before? There's a solution anyone can understand. Those nerds in engineering just confuse everyone with difficult words and "math" that has letters mixed in with the numbers for who knows what.Some months later you've successfully fed all the image assets of your users to a third party LLM, creating a post takes 15 extra seconds to compute, nobody knows the occurence of false positives, and man who said "AI" gets a promotion.
>browsing r9k
>>106573521Relax autismo, AI is actually used for image tagging which is why I brought it up. Admittedly, it's not great at it, but nothing else seems to work that well either way.
>>106573291>But seriously, it never works for prevention because it's infinitely easy to slightly modify an image and then spam it. Surely there's some AI that could be developed to prevent stuff like that?You don't need AI for this. There are already image fingerprinting techniques that are more robust to common image modifications. It's how platforms like YouTube do ContentID to DMCA shit (and they've been doing it long before the current "AI" wave).It's just these are more computationally intensive to generate fingerprints, and compare against (vector searches as opposed to strict equality). So why bother for an Azerbaijani blanket-weaving forum?
