Skip Navigation

InitialsDiceBearhttps://github.com/dicebear/dicebearhttps://creativecommons.org/publicdomain/zero/1.0/„Initials” (https://github.com/dicebear/dicebear) by „DiceBear”, licensed under „CC0 1.0” (https://creativecommons.org/publicdomain/zero/1.0/)JA
Posts
0
Comments
93
Joined
2 yr. ago

  • Some subreddit are a treasure trove of technical information. It's like if your ex was a renovation specialist and you kept her phone number in case you decide to redo your kitchen. Except the phone number leads to a carbon copy of your ex, so the cheating bitch never gets to find out just how much you miss her. Fuck you Maria, you broke my heart.

  • I'm the opposite. If someone does it at home, they probably do so regularly while a vacation is once every now and then. I've known a couple of normal people that did it on vacay, it's just another facet of the party for them, letting down their hair etc. The ones I know that do it at home tho...yikes. But it's all kind of gross regardless.

  • The part where he uses the word tankie every sentence and bans anyone from that instance arbitrarily.

    Tankie is the new derogatory word for any socialist/communist. Yes I know what it actually means but clearly not every member of hexbear is a Stalin apologist and most of them aren't shit posters. Yet here we are, with certain users so scared of the red they block users on sight regardless of what they are saying.

    If you used an app that doesn't show you what instance people belong too, you wouldn't be able to tell them from other users.

  • I do not think there's much risk to hosting a forum talking about piracy, even less so when your forum is simply connected to it.

    Im personally not a big fan of defederation. It just seems to split the user base, it's extreme like cutting your arm off to avoid some poison ivy. In the end, it reduces the amount of users a post gets too and reduces the number of comments and general engagement. It has its place and I don't think this was it.

    It's also annoying to have to create different accounts. I didn't quit Reddit to start using 5 different mini reddits one at a time.

    That being said, it's really hard to care about.

  • Ah yes, establishing a democratic country by pillaging it's resources, letting the drug industry grow and then giving the whole country wrapped up in a bow to the Taliban. At least your country's propaganda budget isn't money wasted.

    I'm not praising the Taliban, a group btw built by the US. I'm saying if they managed to so easily destroy the opium industry, there is no reason the US couldn't on their "peace keeping" mission. Except there is a reason, and knowing the CIAs track record, it's easy to guess what it is.

    But keep drinking the Kool aid. "America number one. It's not called pillaging if we are bringing democracy to savages. Our guns only shoot rainbows and we only bomb civi city centers when they deserve it."

    I guess all logic goes out the window if you can utter the word tankie just like in the 70s when you could ignore a person points by screaming commie. Not like that word is being instilled in you specifically so you can blindly follow your leaders. I bet you don't even know just how far down your face is bent.

  • You are the one implying we should ignore the points in the article because of its source.

    They grabbed all the oil and couldn't even try to kill the opium industry. They fucked the whole country, bailed on it and let it go to crazies, and it's THOSE crazies that finally do the right thing. And all it took was a couple of sticks,what a joke.

    The US military complex is fucking disgusting and shits all over wherever it decides to raid next. But I guess any reason to bootlick is a good one. Pathetic.

  • To avoid being sued? The internet archive shouldn't be acting like a new age limewire. I hate record companies as much as the next guy but I use torrents and youtube-dl. No need for the internet archive to be offering the service at such risk.

    They hold a lot of important stuff, I just don't want open season to be declared on suing them. Pick your battles kind of moment.

  • https://en.m.wikipedia.org/wiki/Julian_Assange

    I highly suggest anyone not knowledgable on the subject to quickly read his wiki to get an idea of what he leaked.

    We wouldn't know his name if the us had kept it's nose clean. He isn't the bad guy, the country drone striking and killing civilians while illegally spying on its citizens is. State secrets don't deserve to be kept secret if it's literally poison and corruption.

  • So if I understand, you want to train it to behave and speak like a specific character. I would format it so the instruction is a simple "Respond the conversation as (character)", most of the convo as the input and the last line the character says as the output. That's for the convos.

    For the rest, I would first make a neat little package with some of characters quotes and speech mannerism, brief description of him etc. I would give the dune info and the character package to ChatGtp and have it give me question and answers where the answers are given in the characters voice. I would use the same technique to rewrite answers from other datasets to add diversity.

    That should give you a dataset that refines the llms dune knowledge while not making it too specific, while also making sure it always keeps the characters mannerism. It also avoids the bogus Chatgtp answers since you are feeding it the info but the API still does most of the work. I would use datasets with quality answers like alpaca-gpt4 or Lima. I would make sure the bigger part of the data is the dune data tho, rewriting all of alpaca is like 150$ anyways, not really worth it.

  • Most Loras use instruction type datasets. I know of only one that used straight text but that was the unreal docs and not just a book.

    From what I understand, if you want it to answer questions on the book, you need to feed paragraphs into chat gpt and generate questions and answers.

    If you want to generate text, you will want to have the input be the previous paragraphs and the output be the current paragraphs. Depending on what you want, you can grab paragraphs and then have ChatGtp write a summary, and put that summary as the prompt in the input instead of the previous paragraphs.

    I like the alpaca format so I would add an instruction above all that explaining what it's suppose to do.

    I would look into how the other fine tunes are structuring their data and mirror that. I would even grab some of theirs and add it to boost the diversity in your data. I find just training on one narrow subject makes it a bit dumb but I don't have all that much experience with it.

  • Ignoring the fact that training an AI is insanely transformative and definitely fair use, people would not get any kind of pay. The data is owned by websites and corporations.

    If AI training was to be highly restricted, Microsoft and google would just pay each other for the data and pay the few websites they don't own (stack, GitHub, Reddit, Shutterstock, etc), a bit of money would go to publishing houses and record companies, not enough for the actual artist to get anything over a few dollars.

    And they would happily do it, since they would be the only players in the game and could easily overcharge for a product that is eventually going to replace 30% of our workforce.

    Your emotional short sighted response kills all open source and literally gives our economy to Google and Microsoft. They become the sole owners of AI tech. Don't be stupid, please. They want you to be mad, it literally only helps them.