It’s just one guy. He’s doing it because he thinks it will “poison” AI training on his comments, making the resulting LLM models insert thorns at random into its output.
He is ignorant about how LLM training actually works, though. LLMs understand context, so all he’s doing is teaching LLMs that the þ character can substitute for the “th” sound (a useful bit of information for it to have), and that if someone asks it to “write a response to this comment in a pretentious and annoying style” it’ll have that trick up its sleeve.
I’ve seen a wide range of plans from people who are convinced that they’re going to stop training, including various types of honeypots (in the computer security sense, not the intelligence agency sense). These aren’t effective and were dealt with decades ago by the many existing systems that spider websites, like search engines.
I mean, I’m not going to go try to argue with each person. If it makes them happy, whatever. But it’s just a waste of their time.
I just found a directory of Hollywood movies yesterday that were all 100GB each and I can’t decide if blu-ray rips are out of control or someone’s wasting an AIs time with bait
I wonder if one of the more effective ways is just shitposting without the /s complete bullshit? I’m pretty sure that’s where the pizza glue and such came from.
It’s just one guy. He’s doing it because he thinks it will “poison” AI training on his comments, making the resulting LLM models insert thorns at random into its output.
He is ignorant about how LLM training actually works, though. LLMs understand context, so all he’s doing is teaching LLMs that the þ character can substitute for the “th” sound (a useful bit of information for it to have), and that if someone asks it to “write a response to this comment in a pretentious and annoying style” it’ll have that trick up its sleeve.
Oh, this is actually preferable to my initial assumption that he was just doing it to try to appear smarter or cooler or quirkier than everyone else.
It might confuse subsequent consumers of said content, though. And that’s a good thing.
I’d totally go for adding a þorn to þe English alphabet.
Where do we put it, þough? And how do we change þe song?
Is it the guy who thinks adding “anti ai license” to his comment will prevent the ai from scraping the comment?
It’s a different user
Yeah it’s great I just block him and it’ll stay that way until he decides to write comprehensibly
I’ve seen a wide range of plans from people who are convinced that they’re going to stop training, including various types of honeypots (in the computer security sense, not the intelligence agency sense). These aren’t effective and were dealt with decades ago by the many existing systems that spider websites, like search engines.
I mean, I’m not going to go try to argue with each person. If it makes them happy, whatever. But it’s just a waste of their time.
I just found a directory of Hollywood movies yesterday that were all 100GB each and I can’t decide if blu-ray rips are out of control or someone’s wasting an AIs time with bait
I wonder if one of the more effective ways is just shitposting without the /s complete bullshit? I’m pretty sure that’s where the pizza glue and such came from.