Keyoxide: aspe:keyoxide.org:KI5WYVI3WGWSIGMOKOOOGF4JAE (think PGP key but modern and easier to use)

  • 0 Posts
  • 158 Comments
Joined 2 years ago
cake
Cake day: June 18th, 2023

help-circle
rss


  • Probably only sucessful ones.
    Google captchas have had multiple rounds (with it faking you out claiming you failed) for probably a decade. Every round of the game updates some confidence score which if you get it high enough lets you pass.
    This conversely means there is no way to fail, you just get stuck in an infinite loop of challenges if your score doesn’t get high enough.

    The only other alternative means of pricing it would see even valid users consume way more than one “verification” per actual completed captcha, since so many users have low enough scores to need multiple rounds of captcha even when completing them with perfect accuracy.
    I doubt they do this, but if they do it’s a scandal waiting to happen, besides also being very weird for any kind of statistic google certainly offers for their captcha.













  • The inputs of the model are full copies of copyrighted data, so the “amount used” is the entirety of the copyrighted work.

    If you want to apply current copyright law to the inner working of artificial networks, you run into the problem that it doesn’t work on humans either.

    A human remembering copyrighted works, be it memorization or regular memory, similarly is creating a copy of that copyighted work in their brain somewhere.
    There is no law criminalizing the knowledge or inspiration a human obtains from consuming media they did not have the rights to consume. (In many places it isn’t even illegal to aquire and consume media you don’t have rights to, only to provide it to others without those rights)

    Criminalizing knowledge, or brains containing knowledge, can’t possibly be a good idea, and I think neural nets are too close to the function of the brain to apply current regulation to one but not the other. You would at minimum need laws explicitly specifying to only apply to digital neural nets or something similar, and it apears this page is trying to work in existing regulation. (If we do create law only applying to digital neural nets, and we ever create intelligent enough ai it could deservedly be called a person, then I’m sure that ai wouldn’t be greatly happy about weird discriminatory regulation applying to only its brain but not that of all the other people on this planet.)

    A neural net is working too similarly to the human brain to call the neural net a copy but the human brain “learning, memorization, inspiration”. If you wanna avoid criminalizing thoughts, I don’t see a way to make the arguments this website makes.






  • Yes, seems you are right. Not sure where I got the impression.

    Unrelated, when I researched this I saw that acme.sh, zerossl, and a bunch of other acme clients are owned by the same entity, “Stack Holdings”/“apilayer.com”. According to this, zerossl also has some limitations over letsencrypt in account requirements and limits on free certificates.

    By using ZeroSSL’s ACME feature, you will be able to generate an unlimited amount of 90-day SSL certificates at no charge, also supporting multi-domain certificates and wildcards. Each certificate you create will be stored in your ZeroSSL account.

    It is suspicious that they impose so many restrictions then waive most on the acme api, where they presumably could not compete otherwise. On their gui they allow only 3 certificates and don’t allow multi-domain at all. Then even in the acme client they somehow push an account into the process.

    […] for using our ACME service you have to create and use EAB (External Account Binding) credentials within your ZeroSSL dashboard.

    EAB credentials are limited to a maximum per user/per day. [This might be for creating them, not uses per credential, unsure how to interpret this.]

    This all does make me slightly worry this block around apilayer.com will fall before letsencrypt does.

    Other than letsencrypt and zerossl, this page also lists no other full equivalents for what letsencrypt does.