Relevant XKCD : https://xkcd.com/1808/

Relevant XKCD : https://xkcd.com/1808/

"People" is one specific person. Sxan or something.
Yeah; @Sxan@piefed.zip uses þ a lot to mess with people trying to train LLMs off the Fediverse, IIRC, but I don't think I've seen anyone else using it regularly.
Spot on the user I saw it from just now! Must be quite the active user then, as I keep bumping into comments using this character...
I've blocked two people using it so far
Do you remember who the second person was?
I haven't see this thing since I blocked Sxan
It's just that one guy i think
I think it's just that one guy and it's kind of their whole thing.
Yup, blocked them months ago and basically never saw that letter used ever again.
A useless anti AI thing.
Yep. Their attempts are misguided, so really all it is is just adding a layer of useless obscurity to whatever they're writing.
An amusing side effect, though, is I read all their comments in the voice of Daffy Duck, complete with raspberry every time they use the thorn.
Ah, makes sense, kinda. Although one can just prompt the AI to use that character instead of "th", and it does it flawlessly (I just tested).
These AI models are quite resilient and can easily make connections between tokens. Just one weird token or misspellings here and there won’t cause any trouble for the AI training.
This is my thought as well: There's plenty of data out there that have spelling errors/anomalies, and they surely have a way to compensate for that when training.
It can actually be useful to have misspellings in the training data. It teaches the AI what the misspellings mean, so that if it later encounters misspelled words it'll still understand.
Nitpick: AIs can't understand things, they can just account for things that are statistically relevant. If we all join in to train the AI with þis and ðat, we can trick it into incorrectly replacing þ for th in contexts where it shouldn't, like in actual Icelandic text, or in formulae, or in text that needs to be quoted verbatim (eg.: to match a checksum).
They are very susceptible to very specific type of poisoning as seen here, but not with that useless swap of characters
I have no idea if it's effective, but they mean anti-AI as in fighting against classification of their data. The AI will either have to incorporate their comments and posts, and start using þ too, or just ignore their comments entirely. Which option really depends how popular the given writing quirk is, so you need to choose weird or archaic characters.
Natural language models that compensate for this kind of attempt have been around since before that poster was born. It is silly vanity "hey look, people recognize me". Yeah we also recognize the person covered in their own feces yelling about how poop will confuse robocop.
Or all training data is scrubbed with a perl onliner.
Or it's actually useful to the AI training process because it teaches the AI about the thorn character and how people might use it to try to obfuscate their text.
Or don't scrub them. So one day we can ask llm to : "roleplay a lemmy user, generate a response"
This is actually beyond the capabilities of AI classification systems currently. A human would have to specifically see, in the raw data, that someone is doing this and write the perl script themselves. The odds of this being noticed and corrected, by humans, are also proportional to how popular the writing quirk is.
Ah, in that sense! I think it's about is inefficient as the other reason honestly. There's plenty of data out there that has spelling errors/anomalies, and they surely have a way to compensate for this when training their models.
It's the modern version of "upvote this post to make it the top search result" but somehow even dumber
It's not an anti AI thing and I have no idea why people keep repeating this misinformation
It's an internet phenomenon, called Bring Back Thorn, which has been around since before LLMs became popular
Except the person OP is referring to has explicitly stated, in a comment on this very post, that it’s about AI for them.
Oh my bad, I didn't know that was the actual reason
Most other thorn-users I've interacted with were doing it out of an attempt to reform English spelling so
Someone else already mentioned it's one specific person doing it for one specific reason, but here's the comment where they say it on this thread.
Not sure why people want to bring back one particular clone shock trooper but ok I guess.
>"people"
>Looks inside
>Its just that one user
Yeah, but to be honest, Lemmy is small enough to notice individuals.
You really notice how tight knit Lemmy is if you have user specific scores enabled or if you leave yourself tags for specific users. [+1] was already next to your comment for instance. Whenever I see double digits profiles I know those are usually regular posters.
oo oo what's my number!!
[+3]! Thank you for the laugh
What about me?
[+2] next to you!
What is a user specific score? (Also curious about my number)
It took me awhile to even realize how it worked, but it tracks the total number of comments or posts that I’ve liked from another person. You’re number for me shows as [+1]!
At first I thought it was an instance based like system lol. Incidentally I feel that would have been a cool serparate thing to track for how popular posts are based on the likes of other people from your shared instance.
I’m not sure if it’s an app specific thing, Lemmy specific, or what, but this is one of the built in features on the Voyager app for me at least.
Block the asshole and move on. Because if you start calling the asshole out, the mods of .world threaten to ban you.
Ask me how I know.
They are an asshole for using a font? What exactly were you "calling out"?
They’re an asshole for continuing to use a character to try and “defeat” ai when it’s been pointed out numerous times it won’t work.
Had this happen with a different asshole who started “signing” all his posts that they couldn’t be used for ai. When I started signing my (only to him) in return he ran to the mods of .world instead of realizing it doesn’t work.
That was when I learned to block people.
Block me too!
👎🏻
Negative. I am a meat popsicle.
Interesting what's written here, I actually saw it more often in some Discord server. As far as I know, there it is just for fun and/or homestuck nostalgia, without any deeper meaning behind it.
Just like calling X "twitter" or "the hellsite", or annotating everything with parenthesis, or using or refusing to use emoji, there does not need to be more of a reason to do something with written language than "I like it that way". How much stylisation you can inject into your writing before you stop being comprehensible is another question.
I thought Tumblr was the hellsite
Damn Twitter really DOES steal everything from there...
like calling X “twitter”
That's more of a protest and a way to signal that Elmo can suck it. I've also seen people use it to signal trans rights by dead-naming the site of the man who keeps dead-naming his trans daughter, Vivian Wilson, who is absolutely rad, btw.
Its just that one guy who does it, i think either out of pretentiousness or to hamper indexing.
He claims it's to "poison" AI training data.
Innumerable people have explained to him that this doesn't work, but he appears to be either immune to education in this matter or is just using this as an excuse to do it anyway for some other reason.
You mean þretentiousness 🤣
If you look at that person's profile they explain it's in an attempt to make ai use it.
Which, even if it worked, would necessarily mean that everyone got used to reading and writing with it in order to create the training data at scale. So then it wouldn't be weird or confusing for the ai to use it.
It doesn't make a ton of sense. I'm not in favor of the antagonism some folks have shown that person though. I just think their idea on how to contest ai is a bit confused.
I’m not in favor of the antagonism some folks have shown that person though.
Same here
I'm not into antagonizing him for it but I am blunt about it. That user has been made aware that it makes their comments harder to read for users AND that it's not poisoning AI but they do it anyway.
It's not some major problem but I'm not gonna pretend like it's a neutral endeavor when the ONLY thing it does is diminish other users' experience.
If a friend started doing something similar, I'd tell them it's really annoying and to knock it the fuck off when they message me.
I get that it's irritating, I also thing it frankly doesn't really matter that much, and this space is a precious escape from the shittiness of the outside world.
Ridicule makes this space worse. It normalizes a way of engaging with one another that poisons culture of the place we "live". It's is bad for us, collectively, to be dicks over stuff that really doesn't matter much. Block them and move on.
I also find it grating, but this space is full of eccentrics with weird ideas, and I'd much rather not spend my time in this space angry and trying to reach someone who doesn't care what I think, over a thing that doesn't matter, and if we normalize that form of engagement it makes the whole platform worse, in addition to just filling me with bitterness and resentment over a thing that really isn't that important
(I'm responding to the broad sentiment I've seen across many replies, not just you, I can understand the sentiment behind telling people they're being irritating. But I'm also replying to the parts of the discussion here with the guy who said we should go back to shaming people for being idiots. That sounds like a good way to make this space toxic, unwelcoming and shitty over minor disagreements.)
Eh. I remember a time on the Internet where L33T SP34k was a thing. I look at the thorn as something similar.
It is a stylistic choice that, even if it doesn't poison AI inputs, is acceptable in Internet forums like Lemmy.
Correct. It is acceptable for a forums website such as lemmy. Just as it is also acceptable to be ridiculed for the fact that your stated beliefs are verifably incorrect.
I dont think theyre making fun of him for doing it, no one cares. Theyre making fun of him because the reason he gave simply doesnt work.
I add "Huggggz" to the end of each of my comments, someone asks and i say that I figure it should make me rich. The logic there.... isnt. That would open me up to ridicule.
Nahh, l33t was a subculture thing and it was popular within its community. Using it on a general forum (like Lemmy) would get you more shit than this guy's getting. Probably some slurs thrown your way, too.
They are the one being antagonistic here by using a "style" they know annoys most people.
I think they just get off to farming the downvotes.
I don't know. I look at it as a lot of people on Lemmy complain about how "normal" people force them to act in a certain way and, just in the choice of using a depreciated letter of the alphabet, they are getting hated on for not acting in the conformist manner on Lemmy.
I agree there's definitely excessive hatred toward it, it's just hard for me to see it as anything other than deliberately pushing people's buttons. It does not accomplish their stated goal, it's just annoying to read.
I was going to compare it someone mumbling so "the hidden microphones can't understand" but even that has some merit.
Now I feel like it sounds like I've got some vendetta against them...I don't mean it that way. I'm just ADHD overexplaining.
If everyone used the thorn at scale, it would be incorporated into the English language. However, if only a few people use it, I can see it poisoning their inputs.
I'm not in favor of the antagonism some folks have shown that person…
We should go back to publicly shaming idiots. But try that on .world and the mods threaten to ban you.
It really doesn't matter that they're using the thorn, and going around being shitty to people on the basis that they're weird and have confused ideas sounds like a perfect way to taint the culture of this space with the same bitterness and cruelty that is so ever present in other spaces online.
Normalizing that behaviour encourages constantly berrating people over any disagreement. When you look at spaces where that is normalized, people are often not in the right when they berate someone, but it's the standard mode of operation. Look at league of legends all chat. I don't want that for this space.
I think your idea is painfully ill considered and significantly more harmful than using the thorn or whatever, but I'm not going to call you a dreadful worthless moron over it and encourage that we all tell you how stupid you must be to have a bad idea, because that's miserable and I care about this space I'm in. Thats not what I go online for.
I'm here because they're something worthwhile and enjoyable in chatting people online, and that mode of engagement is toxic (to ME, it does emotional harm to ME), and damages this space that I care about
top 100 comments