• 0 Posts
  • 1K Comments
Joined 3 years ago
cake
Cake day: June 21st, 2023

help-circle


  • Fair use law is typically up to the judge. The four categories provide you a defense for fair use, but there’s no “hard line”. Unfortunately, “use common sense” is about as good as it gets, at least regarding the law.

    I agree some kind of guideline would be helpful though. Not a hard line at all, but something to get a sense of where it becomes too much.

    Edit: should clarify that I’m not a lawyer, but I’ve seen and read a fair bit of content regarding fair use in the past. It’s actually worth the time to research (if you’re in the US anyway) because you can see just how awful the cases would be if many of the copyright claims we see today were actually taken to court rather than just being big person bullying small person.


  • “I don’t care what adults do [sic], but the books were right there in front of a play mat for children, and for them to say they’re not trying to influence children — then why are they there?”

    Well they are to influence children. Teaching them the real world should be the library’s job, and the real world has families with two daddies, two mommies, or even in the case of one of my friends, two trans mommies and a cis mommy. Educating children influences their worldview in the same way that withholding or miseducating children does too.

    So if you don’t want to influence children, then perhaps you should seek to ban religious material from being given to children. In fact, unlike the books on display, there’s plenty of examples of sexual assault, slavery, murder, and so on in the bible. Let’s restrict that book first so that only adults may read it. Then maybe we can talk about the books where Johnny has two dads and a half-sister.








  • The entropy stems from the words, not characters. With random words and no repetition, you have C(n, k) combinations P(n, k) permutations (or n^k with repetition), where n is your dictionary size and k is the number of words you chain together. These passwords tend to be longer, but not by much. A big enough dictionary can yield some pretty high entropy with only a few words.

    I’ve had passwords as limiting as 16 characters for some services (unfortunately)…

    16 characters is hardly enough for random characters unless you include Unicode (which rarely works for those same services that usually have shitty implementations).

    Not much can be done there, sadly. You’re lucky if they even hash their passwords anyway - you’ll probably just get your password emailed to you if you click “forgot password” like it’s 2003.

    I would never, ever want my users relying on their brain…

    I never would either. People should just use a password manager. I was just mentioning an alternative that generated more memorable passphrases, but I wouldn’t advocate for it over random high-entropy strings saved in a password manager.





  • The README lists the VRAM requirements for their different models if you plan to run with a GPU. Without a GPU, you can translate those roughly to system RAM.

    Note that ML models pretty much always runs faster on a GPU due to the kinds of operations needed to execute them. If you have the option to run on a GPU, you probably should just do that. Even their largest model only requires ~10GB VRAM based on their table, and if you only need English, you can use a smaller one specialized for English (like medium.en).



  • I’m not sure where you live, but here in the US, bats tend to be one of the biggest carriers of rabies alongside various animals that depend on the state. You can actually look up which animals are the primary carriers within a state (if you’re here). Outside of the US, naturally it varies from country to country, but you’d still be looking at bats, plus potentially stray dogs and such. It’ll take some effort, but you should eventually be able to find a carrier.


    I would expect with the nondeterminism and the regular evolution of search results that reproducing the output would be difficult even if DDG didn’t make a patch already. It’s really easy to get these “AI overviews” to make up random shit, though. All you have to do is search normally about a topic that the model wasn’t trained on, and there’s a good chance it’s making shit up, even if only in part of the answer.