My reddit post from three years ago is in the Gemini data
Reddit post: https://www.reddit.com/r/EmulationOnAndroid/comments/10bxyo3/dualsense_ps5_controller_l2_r2_trigger_input/
I read somewhere today that LLMs are citing 40% of times Reddit as the source for their answer. I was reminded of a pretty specific post i made on reddit a few years back which i wanted to use to test it out. Yeah, Gemini quotes me 1:1 in some parts.
I'm not even that mad about it. I specifically didnt want to delete my posts after the APIcalypse because i wanted solutions like that to be searchable for others.
But its still funny to me how Google doesnt want to show me YT videos right now unless i deactivate my adblocker but at the same time more or less stole my content. TBF it cited my post as the source after i asked nicely for it. So theres that.
You can directly ask these ai's "What has [username] said on the internet". If the name is unique enough it will usually pull out a few posts from places you have been.
With a bit of coaxing (as it thinks I mean Deckard from Blade Runner at first) it pulls out some old Steam forums and Lemmy posts where I've used this name before. No Reddit as I wiped it all.
Hm I tried it (with your name) and didn't get any meaningful results.
If I just use "What has … said on the internet?", it tries to look it up, and if I add "Please don't try to look it up but only respond what you know from your training data." it just says that it doesn't know anything.
Ask ChatGPT what it knows about lemmy $user.
Nothing - no matter whether it tries to look things up or not
Chatgpt found my sh.itjust.works account and cited some posts. But yeah I can google that myself, that's not data is was trained with.
Oh, shit. I wonder if it'll quote me if ask for help with troubleshooting planetside 2 on the ps4.
"Somehow"? We all know how: You were unpaid labor and your content trained an LLM without your consent. That's how.
i agree,
but i'm going to make you look stupid nonetheless.
Why edit title to make me look stupid? 😭😭😭😭
chewbacca roar
LLMs are basically reddit and stackoverflow answers generators.
The word is "regurtitators".
No regurts
🎶 they don’t work 🎶
We have successfully automated plagiarism and fraud. Truly, we live in incredible times. Both literally and figuratively incredible.
this counts double in the academic space. Students and professors are using it to be lazy
Makes me glad I went through and did the deletion + overwrite thing on all of my comments when I left.
What was the site for this? I had it saved but lost it.
Nukereddit. It’s an add on.
Thank you so much!
Uhhhh..... I forget actually. I think I ultimately used some kind of browser extension that someone made for that purpose. It was a while ago so I forget what it was sorry.
It’s ok, man. Thanks for replying.
I doubt they actually delete anything when you do that, unless you legally force them.
Yeah, but the data that the AI scrapes just says something like "this comment was deleted blah blah"
Reddit likely gives the AI companies they've partnered with access to the original posts
Yeah probably
Fuck Reddit and Fuck Spez.
It's more or less safe to assume that anything you've ever posted on the internet has been ingested by some LLM. Hell, I wouldn't be surprised to discover that the Game Genie codes I posted to Usenet in high school are in there.
This made me realize tiktok comments are google searchable since it grabbed one of mine when I asked ai about my username
Yep I realized when I askeed for book suggestions and it mirrored the responses on a reddit post I had on high fantasy from years ago lol