>>10174Yes, I'm sure you can recognize posters on a website with only a few dozen posters, or in /mlp/ 4channel threads with only a few dozen posters if that. Can you recognize every poster? Why not? Because there's only a few idiosyncrasies that stand out and not every poster has enough of those to stand out. Most people don't have many of those, and if they do, it only shows in longer writing.
And then, how much will a "tell" stand out if you had a thousand, or ten thousand, or an unknown large number of posters? Suppose a poster has an idiosyncrasy that is present in 10% of posters, and he has three of those idiosyncrasies. Not capitalizing words in the beginning of sentences, not adding an extra space after paragraphs... what have you. That would mean only 1 in 1000 posters would have those three "tells" if each has an occurrence of 10%. I'm sure a 1/1000 chance is good enough to be reasonably sure you're dealing with the same poster if you're on this website, or in a general in /mlp/, but what if you are on an imageboard with 10,000 posters? Out of 10,000 posters, there should be about 10 posters with those three tells. And so on. And did you notice your qualification "after several posts"? Every single thread, you have to start anew trying to connect posts. That means that an AI software would need to start anew every single thread. That's a tiny amount of data to work on compared to, lets say, any social media with handles, which is to say, literally any other form social media besides image boards. And how do you know if you actually have the same person? You never do.
So let's say that someone did train an AI to recognize posters. Okay, congratulations, what do you know about Anonymous1765? Well, he or she doesn't like jews, and lives in the United States if you're posting from a Geo IP and not using a proxy, which the software has no way of knowing... That's not anything meaningful at all. the only way you could possibly be doxxed or seriously lose your anonymity is if you're doing things you already should not be doing. And why would
anyone train a software to do that? It wouldn't work on /pol/ because there's just too many posters and bots already there to make identification possible except for tripfags. There's no reason at all to do it here or on 4channel /mlp/. There's no point and it would be easier to just read the threads yourself. In any case it would be far, far easier to just pay or subpoena Hiro for IP data, which would make the task dramatically easier because then you can compare posting across the website instead of just within the thread. Also, delisting mlpol.net from Google would do nothing at all. You'd only see it in Google if you googled mlpol or similar, and then at that point you may as well just type in "mlpol.net" and you'd get here anyways.
Anyways, it's about as reasonable of a fear as fearing you will get hit by a meteorite in your room... except that actually has a non-zero chance of happening. Just don't be a dumbass when you post online, and don't dox yourself.