You must log in or # to comment.
Does Reddit not persist post and comment revision history? If they do-- as a developer imaging myself in charge of such a feature-- I would just use full post and comment revision history for training, directly from the database.
This extension probably feels great, but may accomplish very little.
They do. After I deleted all my comments, using automatic tools (that replace the text) and manually, they keep recreating them. In fact, this might sound a bit like a conspiracy, but I’ve noticed all my comments that do come back are the ones that people find coming from Google.
So everything is deleted, then some user searches Google for a solution and my comment was the only one, as soon as they click the post, my comment is back and shows up in my account. The original comment, not even the modified version that should’ve replaced it.
So 100% Reddit keeps everything.
Then the solution is to continue posting on reddit and poisoning your own post with random nonsense words and hope that does something, I guess.
setting down the molotov cocktail and lighter oh, we’re doing it the “legal” way, apparently
I would never advocate for anyone to burn down reddit. It’s a ridiculous proposal. Nothing would be gained if you were to throw molotov cocktails on the reddit servers.
Absolutely not. Do not use reddit. Its time has long past.
It’s always easy to imagine why things won’t work, then decide to do nothing. Sometimes just getting your opponent to respond opens new opportunities, while predicting failure such that you don’t bother is a guaranteed path to defeat.
Also, just because they kept previous revisions doesn’t necessarily make it pointless, because they’d still be using the edit in the training data. And yes, they can probably figure out how to clean that up, but then let’s make them do that and see what happens.
edit: Most importantly, shit posting is fun.
Yeah they’ll be able to train a LLM to look out for, and filter out of future training, off-topic, sabotaged posts.
Lol the part about non-copyrighted text definitely should be read with a wink.
You can use any text that you want, but please, do not choose something copyrighted. The New York Times is currently suing OpenAI for training ChatGPT on its copyrighted material. Reddit’s data is uniquely valuable, since it’s not subject to those kinds of copyright restrictions, so it would be tragic if users were to decide to intermingle such a robust corpus of high-quality training data with copyrighted text.
😇
I have mixed feelings about this… Reddit was an incredible source of knowledge and now it feels the the Library of Alexandria is burning down.
I would much rather see an extension that copies your comments off of Reddit and onto another location… Ideally into an open source LLM model.
Yeah, would be awesome if you could move them to a Lemmy community and then point to that. Then replace it with “I’ve moved to the Fediverse and so should you. To see this comment, follow this link.”
Yeah, me too. Here’s how I think about it, though: The French are famously proud of Paris. They love it. The French government also knows that if they push their citizens too hard, they will burn Paris to the ground. This is, surprisingly, very healthy, and has allowed the French to resist the neoliberalization that has swept the rest of the west much more successfully. Meanwhile, Americans would never do such a thing, so we don’t get healthcare, pensions, vacation days, etc. Tech companies are insufficiently afraid of their users. They should know that we’ll burn down the internet should they displease us. We might end up losing a few valuable things in the short term, but in the long term, we’ll have a much healthier relationship.
